Application Of Quantitative Magnetic Resonance Imaging In Alzheimer’s Disease

Part 1Application study of machine learning models based on arterial spin labeling combined with T1 mapping for the diagnosis of Alzheimer’s diseasePurpose:Arterial spin labeling(ASL)combined with T1 mapping was applied to explore the changes of cerebral blood flow(CBF)and T1 values in Alzheimer’s disease(AD)and mild cognitive impairment(MCI).Methods:A total of 34 AD patients,24 MCI patients and 38 normal controls(NC)were included in this study.The ASL sequences were integrated into magnetizationprepared 2 rapid acquisition gradient echo(MP2RAGE)sequences to obtain CBF and T1 values for 24 brain regions automatically segmented.T test and U test(Bonferroni correction)were used to select features with significant differences among groups.Two feature screening methods,(recursive feature elimination and Relief)and two classifiers(support vector machine and logistic regression)were used to construct the best diagnostic model.Receiver operating characteristic(ROC)curve and decision curve analysis(DCA)were used to evaluate the diagnostic performance of the model.The consistency between predicted and actual probabilities of the model was assessed using calibration curves,and the Hosmer-Lemeshow test was used to assess the goodness-of-fit of the model.Results:A total of four imaging features,left insula T1 value,left hippocampus T1 value,right insula T1 value,and left hippocampus CBF value,and one clinical feature(age)were screened,and five models were constructed,including:age,T1,ASL,T1+ASL,and T1+ASL+age.For the discrimination of AD and NC,T1+ASL+clinical data had the best diagnostic efficacy(AUC=0.931);for the discrimination of MCI group and NC,the AUCs of ASL model,ASL+T1 model,T1+ASL+clinical model,and T1 model were 0.793,0.764,0.773,and 0.709,respectively,and none of the differences were statistically significant(P>0.05);for AD and MCI discrimination,the ASL-only model AUC=0.531 was lower than the T1 model(AUC=0.724)and the ASL+T1 model(AUC=0.727),and the difference was statistically significant(P<0.05),while the ASL+T1+clinical model did not show higher diagnostic efficacy(AUC=0.674).the T1+ASL model calibration curve showed better predictive accuracy between actual and predicted probabilities.Conclusion:The ASL technique combined with T1 mapping technique can distinguish NC,MCI and AD than the two techniques alone.This shows that ASL technique combined with T1 mapping technique may be used as a powerful tool for accurate diagnosis and disease course monitoring of AD.Part 2Application of arterial spin labeling combined with amide proton transfer in Alzheimer’s diseasePurpose:The arterial spin labeling(ASL)technique combined with amide proton transfer(APT)technique was used to analyze the differences in specific brain region of Alzheimer’s disease(AD)and evaluate its diagnostic efficacy.Methods:29 AD patients and 19 normal control(NC)attending our hospital from September 2020 to December 2022 were enrolled.Magnetization-prepared 2 rapid acquisition gradient echo(MP2RAGE),ASL and APT images acquired at 3.0T MRI.The cerebral blood flow(CBF),MTRasym and AREX in 24 brain regions were obtained on Matlab platform automaticlly.The differences in CBF,magnetization transfer ratio asymmetry(MTRasym)and apparent exchange dependent relaxation(AREX)between the two groups were compared by independent sample ttest(Bonferroni correction,P<0.05 after correction).The diagnostic efficacy of CBF single-parameter model,MTRasym single-parameter model,AREX single-parameter model and combined models was evaluated by ROC curves.DeLong test was performed to evaluate the diagnostic performance among different models.The correlation between CBF,MTRasym and AREX with the mini-mental state examination(MMSE)was observed by Pearson correlation analysis.Results:The CBF of six brain regions,including left hippocampus,left occipital gray matter,right occipital gray matter,left insula,left temporal gray matter,and right temporal gray matter were lower in the AD group compared with the NC group(P<0.05).The MTRasym values were not different in the AD and NC groups(P>0.05).The AREX values of three brain regions,including left occipital gray matter,right occipital gray matter,and right insula,were lower in the AD group compared to the NC group(P<0.05).The area under curve(AUC)of the single parameter CBF models for differentiating between AD and NC include the left hippocampus,left occipital gray matter,right occipital gray matter,left insula,left temporal gray matter,and right temporal gray matter were 0.797,0.757,0.753,0.742,0.733,and 0.759,respectively.While the AUC for the combined CBF model was 0.786.The area under the ROC curve of the single parameter AREX models for identification of AD and NC include left occipital gray matter,right occipital gray matter,right insula were 0.764,0.710,0.723,respectively.While the AUC for the combined AREX model was 0.744.The AUC for combined CBF in six brain regions and AREX in three brain regions was 0.900.CBF in the left hippocampus,left occipital gray matter,right occipital gray matter,left insula,left temporal gray matter,and right temporal gray matter were positively correlated with MMSE scores(r=0.808,0.583,0.687,0.625,0.702,0.687,respectively).The AREX values of left occipital gray matter,right occipital gray matter,and right insula were not correlated with MMSE scores(P>0.05).Conclusion:The ASL technique and APT technique can provide information on different pathological changes in the brain of AD patients,and the combination of CBF and AREX can better distinguish AD and NC.Thus,the ASL technique combined with the APT technique may serve as a powerful tool for the diagnosis of AD.