Research On Alzheimer’s Disease Of The Senior Citizens In Communities In Nan Chang Based On Artificial Neural Network Model

Objective: The purpose of this study was to evaluate the characteristic and regularity of Alzheimer’s disease(AD)among senior citizens in urban communities and to analysis the risk factors.In addition,neural network fit for senior citizens in urban communities model will be built to realize the early diagnosis of AD and provide scientific thoughts for the prevention.Methods: Nested case-control study applied in this study,89 new cases with good compliance who are willing to provide blood and urine were sampled from the follow-up cohort in preliminary work with 198 non-AD controls(1:2)living nearby.Questionnaire investigation and laboratory experiments were used to analysis the risk factors and biological indicator.Moreover,Error Back Propagation Artificial Neural Network was built on the basis of the data collected.Results:(1)In Nested case-control study,it shows significant difference(p<0.05)that the average age of cases(77.44±6.82)was higher than the controls(72.49±6.86)and the proportion of women in AD group(69.66%)was higher than that in non AD group(49.44%).(2)after adjusting for age and gender,the average score of MMSE scale in the AD group was(17.64 ± 5.38),which was significantly lower than that in the control group(26.57 ± 3.63),t=21.94,p<0.001.The average content of urine AD7C-NTP in AD group is(1.99 + 1.94)ng/ml,which is still significantly higher than the control group(0.96 + 1.31)ng/ml(t=5.57,p<0.001).Its area under the ROC curve was 0.851(95% confidence interval 0.794-0.907,p<0.001).The optimal cut-off point of AD7C-NTP was 1.374ng/ml,and the corresponding sensitivity and specificity were the highest,which were 77.3% and 87.6% respectively.However,there was no significant difference in plasma Aβ42、Aβ40 or Aβ42/Aβ40 between cases and controls(p>0.05)after adjusting.(3)The built ANN model shows that three scale scores accounted for the highest weight(all more than 0.1),of which the highest is MMSE scale(more than 0.2),followed by is the Mo CA scale(0.1167),the weight of ADL scale is 0.1051.In addition to the scale score,age is the most important factor(the weight is 0.0775),followed by is the biological index: urine AD7C-NTP(the weight is 0.0693).The weight of family history,education level,occupation,physical exercise and physical labor were all higher than 0.05,which ranks in front of other multi-dimensional factors.(4)After 270 times learning,the built ANN model was in a stable state and its error reached the lowest(MSE=0.0112988).Its accuracy of AD diagnosis was 96.6%,and the area under the ROC curve was 0.947(95%CI = 0.903-0.991,p<0.001).Conclusion:(1)The incidence risk of AD among senior citizens in urban communities is highly correlated with age,and female patients are relatively more than males.Family history of dementia,educational level,occupation,physical exercise,family income and other factors in many aspects all have a certain influence on the incidence of AD.(2)In biological indexes,the content of urine AD7C-NTP between patients with AD and healthy people is significantly different,which is valuable in clinic for early diagnosis of AD.However,the plasma Aβ42、Aβ40 and their ratio are not associated with AD,and the clinical significance of them needs further investigation.(3)The built ANN model for AD diagnosis among senior citizens in urban communities has a good accuracy and diagnostic efficiency,and is economical,which can be used for AD screening in a large population.