Application Of Resting-state Functional Magnetic Resonance Imaging And Neurocognitive Scale In HIV-related Neurocognitive Impairment

PART ONE: FEASIBILITY STUDY OF MINI-MENTAL STATE EXAMINATIONIN DIAGNOSIS OF HIV-ASSOCIATED NEUROCOGNITIVE DISORDERS OBJECTIVE: Taking neuropsychological test as gold standard,it is one of the methods for clinical diagnosis of HAND to study whetherMini-mental State Examination Scale can replace the neuropsychological test.METHODS: According to inclusion exclusion criteria,29 AIDS patients(HIV-positive group),16 males and 13 females who were not treated with antiretroviral therapy were recruited from the infectious department of the Fourth People’s Hospital of Nanning from September 2007 to December 2008.At the same time,this study recruited 20 age-and gender-matched normal controls(HIV-negative group),12 males and 8 females.The subjects were right-handed,and the patients were in asymptomatic stage of HIV infection.Mini-mental State Examination(MMSE)and Neuropsychological Test(NP)were used to evaluate the neurocognitive function of HIV-positive and HIV-negative groups respectively.The correlation between the two cognitive assessment scales was analyzed by SPSS20.0 statistical software.The diagnostic and screening abilities of the two scales for HIV-associated neurocognitive disorders were compared.RESULTS:(1)There were significant differences in MMSE between HIV-positive group and control group;(2)Neuropsychological tests showed that there were significant differences in five tests: digital symbol,connection test,conceptual fluency,vocabulary fluency,Stroop-C and Strop-CW between HIV-positive group and HIV-negative group(p < 0.05).In the two tests of digital breadth and Wisconsin card classification,there were significant differences in MMSE between HIV-positive group and HIV-negative group(p < 0.05).There was no statistical difference(p > 0.05);(3)The correlation between MMSE and NP test was low(r = 0.485,P < 0.01);(4)The ability of NP test to diagnose HAND was better than that of MMSE.CONCLUSION:MMSE fails to meet the criteria of NP test for diagnosis of HAND,and can not be used as a substitute for NP test for diagnosis of HAND.PARTTWO: DIAGNOSTIC VALUE OF RESTING-STATE FUNCTIONAL MAGNETIC RESONANCE IMAGING BASED ON MACHINE LEARNING IN HIV-ASSOCIATED NEUROCOGNITIVE DISORDERS OBJECTIVE: Using resting-state functional magnetic resonance imaging(fMRI)based on machine learning method,the application value of regional homogeneity and amplitude of low frequency fluctuation in evaluating HIV-related neurocognitive impairment was explored.METHODS: The object of study is included in the exclusion criteria as in the first part.Subjects were asked to sign the informed consent form for magnetic resonance examination and perform resting-state functional magnetic resonance scanning.Conventional axial T1 WI,axial T2 WI,sagittal T1 WI thin slice scan and gradient echo-echo plane imaging sequence(EPI)were performed with GE Discovery MR 750 W 3.0T.SPM12 and DPARSF3.1 are used for image preprocessing,ReHo,ALFF and fALFF are used for formal processing,and PRoNTo2.1.1 machine learning toolkit is used for statistical analysis of image data.SPSS20.0 statistical software was used to analyze the differences of Reho,ALFF and fALFF in HAND group and non-HAND group.RESULTS:(1)ReHo index contributed the most to the difference between HIV-positive and HIV-negative groups in ten brain regions in turn: left paracentral lobule,right paracentral lobule,right cerebellum 7b,right motor supplement area,left motor supplement area,right superior occipital gyrus,left superior occipital gyrus,right inferior frontal gyrus,right cuneiform lobe and right posterior central gyrus;(2)ALFF index contributed the most to HIV-positive group and HIV-negative group.The ten regions with the greatest contribution among negative groups were left paracentral lobule,right paracentral lobule,right cuneiform lobe,left middle temporal gyrus,left anterior cuneiform lobe,vermis 1,2,right complementary motor area,left angular gyrus,left parietal gyrus and vermis 10;(3)fALFF index contributed the most among HIV-positive and HIV-negative groups.The brain regions were left paracentral lobule,right cerebellum,left cerebellum,right paracentral lobule,left motor supplement area,left superior temporal gyrus,cerebellar vermis 9 area,left superior frontal gyrus,right medial superior frontal gyrus and right superior parietal gyrus in turn;(4)The results of classification by three indexes showed that the AUC of local consistency(ReHo)was 0.69,and the accuracy rate was 5.9.18%,sensitivity 79.31%,specificity 30%,positive predictive value 62.16%,negative predictive value 50%.The AUC of low frequency amplitude(ALFF)was 0.65,the accuracy was 63.27%,the sensitivity was 75.86%,the specificity was 45%,the positive predictive value was 66.67%,and the negative predictive value was 56.25%.The AUC of low frequency amplitude fraction(fALFF)was 0.62,the accuracy was 61.22,the sensitivity was 75.86%,the specificity was 40%,the positive predictive value was 64.71%,and the negative predictive value was 53.33%.Reho values of HAND group and non-HAND group were significantly different in left paracentral lobule and right paracentral lobule,and ALFF values were significantly different in right paracentral lobule,vermis 1,2 and 10 brain regions.CONCLUSIONS: ReHo,ALFF and fALFF indicators based on machine learning have significance in the evaluation of cognitive function in HIV positive and negative groups.ReHo and ALFF indicators can be used to diagnose HAND,and the diagnostic efficiency and other classification evaluation efficiency of ReHo and ALFF indicators are better than fALFF.The brain regions with significant difference between HIV positive group and HIV negative group included bilateral paracentral lobule,motor assist area,cerebellar hemisphere,cerebellar vermis and part of frontal,temporal,parietal and occipital lobes.Resting-state functional magnetic resonance imaging based on machine learning method can diagnose HAND more comprehensively and objectively than NP test.