Prediction Model Of Early Alzheimer’s Disease Based On Proteomics And Mri Brain Image Texture

Background Alzheimer’s disease(AD)is a chronic neurological degenerative disease,the clinical manifestations of the disease for the memory of the degradation and cognitive dysfunction,a serious threat to the quality of life of the elderly and life safety The Europe and the United States in the elderly over the age of 65 the prevalence of dementia is about 4 to 8%,the prevalence of dementia in China was 7.8%,of which the prevalence of AD was 4.8%.As the world’s most populous country,China will face increasingly serious population aging,AD will give patients,families and society a great heavy economic and living burden.AD from the emergence of clinical symptoms to the initial diagnosis of the average time is greater than one year,and the condition is mostly moderate(67%),early diagnosis of AD research has been one of the hot and difficult problems at home and abroad.The development of AD involves a small and complex change in the structure of the brain,and it is potentially valuable to predict the progress of AD through the microscopic texture feature of the image.The use of Magnetic Resonance Imaging(MRI)has been shown to have a significant effect in predicting the conversion of mild cognitive impairment(MCI)to AD and declining cognitive function in the elderly.Similarly,plasma proteomics has been shown to have a valuable value in diagnosing AD and predicting the conversion of MCI to AD.Combined with plasma proteomics and MRI imaging as biomarkers have potential advantages in early AD diagnosis and prediction.Gaussian Processes(GP)classification shows a strong ability in MRI studies of models for clinical disease identification or prediction.Gaussian process is based on statistical learning theory and Bayesian theory developed a supervised machine learning algorithm,Gaussian process generalization ability,ultra-parameter setting flexibility,with non-parametric inference and probability output,etc.,for dealing with non-Linear and high dimensional and other complex regression problems.Objective OBJECTIVE To establish the early prediction model of Alzheimer’s disease by Gaussian process,and to provide evidence for early diagnosis of AD by extracting the microscopic texture features of brain images based on Contourlet transform and combining the plasma proteomics biomarkers.Methods A total of 420 data were collected from this study,including 84 patients with AD,287 patients with MCI and 49 patients with normal control.The hippocampus region was segmented from the brain MRI images of the coronal bits using the regional growth method.The image of the hippocampal region was processed by Contourlet transform and 14 texture values were calculated.Based on the baseline data,t-test or variance analysis was used to compare the differences between groups.The variables between the plasma protein groups were statistically significant by LASSO(Least Absolute Shrinkage and Selection Operator).Then,the Gaussian process model And the support vector machine model to classify the model,consider the combination of kernel function to select the best classification model and do cross validation;based on MCI patients baseline basic information,plasma protein data and brain image data,follow-up period is converted to AD as the outcome label Modeling and establishing early AD classification prediction model.Results There were significant differences in the plasma levels of Apo AII,FSH and FASLG receptor between the AD and the healthy control group.The AD group and the healthy control group were followed up for 1 year,There was a significant difference in plasma protein between groups.LASSO regression analysis showed that 20 kinds of plasma proteins could be used as potential biomarkers for early AD diagnosis.The sensitivity was 76.2% and the specificity was 81.3%.The area under the ROC curve was 80.4%(95% CI: 86.2% ~ 79%).The MCI group was converted to AD for outcome,and the plasma proteins with differences between groups were LASSO regression and the sex and age were corrected to obtain BNP IL16,TBG,APOE,PLGF,TFF3 and other six kinds of plasma protein,the sensitivity of 91.2%,specificity of 78.4%,ROC curve area of 84.1%(95% CI: 91.8% ~ 81.6%).Based on the basic information of the study object,the Gaussian process classification model was established based on left and right measurements and bilateral hippocampal brain image texture features.Based on the classification model of AD and healthy control group,the sensitivity of the right hippocampus was 91.2% and the specificity was 81.6%,which was larger than that of the left hippocampus.The area under the ROC curve based on bilateral hippocampus was larger than that of The left and right hippocampal images are modeled separately(0.851 and 0.901).Based on the prediction model of MCI baseline data and follow-up outcome,the accuracy rate of MCI transformation is 88.4%,and the accuracy rate of MCI is 80.0%.SVM model predicts the accuracy rate of MCI conversion to 81.0%,and the accuracy rate of MCI return to normal is 60.0%.ConclusionsThe combination of plasma protein levels of IL-16,TBG,BNP,TFF3,PLGF and ApoE distinguishes between AD patients and healthy individuals that can be used for early diagnosis and monitoring of AD and prediction of MCI conversion to AD;a combination of Gaussian radial basis functions The prediction function of Gaussian process is better.Based on MR image texture and prediction model of plasma protein data,it has a positive effect on early prediction of AD.