Altered Spontaneous Brain Activities In Maintenance Hemodialysis Patients With Cognitive Impairment And The Construction Of Cognitive Function Prediction Models

Objective:To evaluate the spontaneous brain activity changes of maintenance hemodialysis(MHD)patients with cognitive impairment(CI)based on resting functional magnetic resonance imaging,and to establish a predictive model of cognitive function in maintenance hemodialysis patients by combining spontaneous brain activity changes with clinical indicators.Methods:From September 2020 to December 2021,50 patients undergoing maintenance hemodialysis in the Second People’s Hospital of Changzhou Affiliated to Nanjing Medical University were enrolled;Twenty-eight healthy volunteers were recruited during the same period,and all subjects underwent neuropsychological testing and magnetic resonance scanning.MHD patients were divided into MHD-CI and MHD-NCI groups based on neuropsychological test scores.Use the least significant difference method to assess the differences between sets of data.P<0.05 was considered statistically significant.After image preprocessing,data analysis was performed to obtain Amplitude of low frequency fluctuation(ALFF),fractional amplitude of low-frequency fluctuations(f ALFF)and regional homogeneity(Re Ho).Use the DPABI toolbox to calculate differences between groups;Postmortem testing is performed using ANOVA with least significant difference,and Gaussian random field correction(GRF)is performed using Z-mapping to correct for multiple comparisons.The Z-score of the data was standardized using MATLAB software,and the Pearson correlation between m ALFF/mf ALFF/m Re Ho values and clinical variables and MOCA scores was analyzed with significant differences between MHD-CI patients and MHD-NCI groups,and clinical indicators related to cognitive function and spontaneous brain activity levels were included in the backpropagation neural network(BPNN)to establish the best cognitive function prediction model.Results:Compared with the MHD-NCI group,patients with MHD-CI had lower haemoglobin levels and higher urea nitrogen levels(96.16±15.73 g/L vs 112.95±9.25g/L,P<0.001;30.69±6.17mmol/L vs 21.19±7.39mmol/L,P<0.001).Compared with the MHD-NCI group,patients with MHD-CI had lower left posterior gyrus m ALFF values(P<0.001),lower left inferior temporal gyrus mf ALFF values(P<0.001),and higher m ALFF values in the right caudate nucleus(P<0.001).Correlation analysis showed that hemoglobin was positively correlated with MOCA score(r=0.640,P<0.001),urea nitrogen level was negatively correlated with MOCA score(r=-0.489,P<0.001),left posterior gyrus m ALFF value was significantly positively correlated with MOCA score(r=0.457,P=0.001)in patients with MHD,and left inferior temporal gyrus mf ALFF value was positively correlated with MOCA score(r=0.295,P=0.038).The right caudate m ALFF value was negatively correlated with the MOCA score(r=-0.455,P=0.001).Based on different correlation factors,four different BPNN prediction models were constructed,and the results showed that the model that included the three related factors of left posterior gyrus m ALFF value,hemoglobin and urea nitrogen had the best diagnostic effect,and the mean squared error(MSE),root mean square error(RMSE),mean absolute error(MAE),mean absolute percentage error(MAPE)and R-squared value(R~2)between the actual score and the predicted score were 0.5840,1.8468,1.7424,respectively.7.78%and 0.8050.Conclusion:In summary,the left inferior temporal gyrus and left postcentral gyrus might be the critical regions affecting cognitive function in MHD-CI patients,and correction of anemia and adjustment of urea nitrogen levels might help prevent CI in MHD patients.Combined with rs-f MRI not only reveals the neurophysiological mechanism of cognitive impairment,but also can serves as a neuroimaging marker for the diagnosis and evaluation of cognitive impairment in patients with MHD.