A Study On Cognitive Dysfunction And Machine Learning Classification Of Obstructive Sleep Apnea Based On Rs-fMRI

Purpose:Obstructive sleep apnea(OSA)patients often have cognitive impairment,and it is closely related to Alzheimer’s disease(AD),which is one of the important influence factors of AD.Preventing the development of OSA towards AD has important social significance and economic value,while the precursor stage of AD,mild cognitive impairment(MCI),is a bridgehead for sniping at AD.Preventing the development of OSA without MCI(OSA-n MCI)into OSA-MCI(OSA-MCI)is crucial.However,the neuroimaging mechanisms underlying the transition from OSA-n MCI to OSA-MCI remain unclear Method:In this study,26 OSA-n MCIs(36.15 ± 9.48),26 OSA-MCIs(39.26 ± 10.25),and26 healthy controls(39.84 ± 11.18)matched to their age and gender were recruited from the Sleep Monitoring Room of the First Affiliated Hospital of Nanchang University,and their resting functional magnetic resonance data and clinical related data were collected.We used the sliding time window method to analyze the differences in dynamic low-frequency amplitude variability among the three groups,and analyzed the relationship between it and clinical indicators.Result:Compared with OSA-MCIs,the d ALFF values of the posterior cerebellar gyrus and right superior frontal gyrus in OSA-n MCI group increased;Compared with HCs,d ALFF values in the right posterior cerebellar lobe of OSA-n MCI patients increased.In OSA-n MCI patients,there was a positive correlation between bilateral posterior cerebellar lobe and right superior frontal gyrus d ALFF values(r=0.971,p<0.001;r=0.943,p<0.001,r=0.943,p<0.001);However,in OSA-MCI patients,a positive correlation was observed only between the bilateral posterior cerebellar lobes(r=0.785,p<0.001).In OSA-n MCIs,the d ALFF value of the left posterior cerebellar lobe was positively correlated with the apnea hypopnea index(r=0.457,p=0.019),the epworth somnolence scale score(r=0.553,p=0.003),and the arousal index(r=0.475,p=0.014);The d ALFF value in the right posterior cerebellum was positively correlated with the apnea hypopnea index(r=0.415,p=0.035)and negatively correlated with the lowest oxygen saturation SO2(r=-0.410,p=0.037).Conclusion:This study suggests that HCs,OSA-n MCIs and OSA-MCIs exhibit different temporal variability in dynamic brain function,and OSA-n MCI may have variable intermediate states.We conclude that dysfunction of the cerebellar prefrontal cortex pathway in OSA-MCIs may lead to cognitive impairment in OSA.Purpose:The purpose of this study is to develop and validate disease classification methods for obstructive sleep apnea through functional activity and connectivity using radiomics analysis.Method:110 clinically confirmed OSA patients(38.24± 9.70)and 122 healthy controls(39.65 ±9.73)underwent resting state functional magnetic resonance imaging(rs-f MRI).After preprocessing,a total of 7134 features of 5 categories were extracted,including z-conversion regional homogeneity(z Re Ho),z-conversion low-frequency fluctuation amplitude(z ALFF),z-conversion ratio low-frequency amplitude(zf ALFF),z-conversion resting state functional connectivity(z RSFC),and z-conversion degree of centrality(z DC).Then,the predicted features were selected through Mann-Whitney U test and variables with high correlation were removed.The minimum absolute contraction and selection operator(LASSO)method is further used to select features.Finally,support vector machine(SVM),extreme gradient boosting(XGBoost),and artificial neural network(ANN)were used to construct classification modeling,and the differences in model performance among the three groups were compared.Result:48 features were selected,including 15 z RSFCs,11 z ALFFs,10 z ALFFs,10 m DCs,and 6 z Re Hos.Based on these 48 features,the accuracy and area under curve of the model constructed using SVM,XGBoost,and ANN in the training dataset were89.19% and 0.95,78.38% and 0.93,83.78% and 0.94,respectively,while the accuracy and AUC of the model in the validation dataset were 76.60% and 0.85,70.21% and 0.83,76.60% and 0.85,respectively.Conclusion:These findings indicate that the effective radiomic method of rs-f MRI can identify OSA individuals from healthy controls with high classification accuracy,providing a potential auxiliary method for clinical diagnostic systems.