Application Research Of Machine Learning Method Based On MRI Structure Imaging And Resting State Function Imaging Image Characteristics In AIDS Patients

PART ONE: THE RELATIONSHIP BETWEEN THE DEGREE OF COGNITIVE DISORDER AND THE SEVERITY OF INFECTION IN AIDS PATIENTSOBJECTIVE: To study the correlation between the degree of cognitive impairment and the severity of infection in AIDS patients.MATERIALS AND METHODS: According to the exclusion criteria,29 new cases(16 males,13 males)who were not treated with antiviral therapy in the Department of Infectious Diseases of the Fourth People's Hospital of Nanning during the period from September 2009 to January 1919 were collected.Females,age-and sex-matched normal volunteers in 20(12 males,8 females)were included in the study.Eight out of 49 subjects with brain atrophy,arachnoid cyst,intracranial infection,and excessive motor artifacts were excluded,and finally included in the AIDS group of 27 patients(14 males,13 females,age range 22-63)The mean age was 42.48±13.03 years old;14 of the normal volunteers were included in the control group(8 males,6 females,age range 22-63 years,mean age 39.0±13.02 years).The laboratory test indicators(peripheral blood CD4+ T cell count and CD8+ T cell count)were collected from 27 patients;the subjects were scored using the Simple Intelligence Scale(MMSE)and Neuropsychological Assessment(NP).In the test,AIDS patients were divided into cognitive impairment according to the scores and test results;22 patients with normal cognitive function and 5 patients with cognitive impairment(including 2 cases of mild cognitive impairment and 3 cases of moderate cognitive impairment)),while the control group's simple smart scale test showed normal cognitive impairment.In the AIDS group,there were 9 cases in the UN group,6 cases in the ANI and MND groups,and 12 cases in the HAD group.Using spss19.0 statistical software,the patient's infection level(CD4+T cell count,CD4+T and CD8+ T cell count ratio)was correlated with cognitive grouping(MMSE method,NP method)and clinical scale.Multiple regression analysis was performed using CD4+ T cell count,CD4+ T cell count and CD8+T cell count as the dependent variable and each neurometric assessment as independent variables.RESULTS:(1)There was a moderate negative correlation between the completion time of Stroop C and the CD4+ T cell count in each AIDS group(P=0.004,r=-0.535),obtained by NP method of each AIDS group.There was a moderately negative correlation between cognitive grouping and CD4+ T cell counts(p=0.02,r=-0.444).(2)There was a moderate negative correlation between the time of completion of Stroop C and the CD4+ T cell count/CD8+ T cell count in each AIDS group(P=0.034,r=-0.409).The NP method of each AIDS group showed a moderate negative correlation with the CD4+ T cell count/CD8+ T cell count(P=0.005,r=-0.527).(3)In the multiple regression analysis of CD4+ T cell counts,the word color interference experiment(SC time)and the lexical fluency experiment(group words)entered the regressionequation.In the multiple regression analysis of the ratio of CD4+ T cell count to CD8+ T cell count,the color interference experiment(SC time)entered the regression equation.CONCLUSION: There is a correlation between the degree of cognitive impairment and the severity of the disease in the AIDS group.Among them,the color interference experiment(SC time)is related to the severity of infection.PART TWO:APPLICATION OF MACHINE LEARNING METHOD IN THE MEASUREMENT OF GRAY MATTER VOLUME IN AIDS PATIENTSOBJECTIVE: To explore the diagnostic value of gray matter volume in the assessment of HIV-related neurocognitive disorders.To explore the correlation between the gray matter volume of each brain region and various clinical indicators using magnetic resonance imaging based on machine learning.MATERIALS AND METHODS: The inclusion criteria for inclusion in the study were the same as in the first part.All subjects were right-handed and signed an informed consent form for magnetic resonance imaging using a GE Discovery MR 750 w 3.0T magnetic resonance scanner.A conventional axial T1 WI scan,an axial T2 WI scan,and a sagittal T1 weighted(Sag3D T1WI-BRAVO)thin layer scan were performed.Image preprocessing used two software,SPM12 and DPARSF3.1,to formally process the PRoNTo2.1.1Machine Learning Toolkit and DPARSF3.1 to extract gray matter volume values and image data analysis.The SPSS19.0 statistical software was used to analyze the difference of the gray matter volume index of 90 brain regions in the HAND group and the non-HAND group.SPSS19.0 statistical software was used to analyze the correlation between gray matter volume value of the top ten brain regions and clinical hematology,clinical scale and cognitive scores and groupingin the normal group and the AIDS group.RESULTS:(1)The ten brain regions with the greatest contribution to the difference in gray matter volume between AIDS patients and normal controls in machine learning were the right postcentral gyrus,the left superior parietal gyrus,the right paracentral lobule,the right supplementary motor area,the left inferior parietal gyrus,the left heschl gyrus,the right inferior parietal gyrus,the right superior parietal gyrus,the right rolandic operculum,and the left supramarginal gyrus.(2)The ROC value of the gray matter volume classification effect evaluation was 0.73,the accuracy rate was 70.73%,the sensitivity was 85.19%,the specificity was 42.86%,the positive predictive value is 74.19%,and the negative predictive value is 60.00%.(3)The gray matter volume index was meaningful for distinguishing between the HAND group and the non-HAND group.(4)The gray matter volume index of the top ten brain regions had different degrees of correlation with clinical hematology,clinical scale and cognitive scores,and grouping.CONCLUSION: Machine learning is useful for the classification of gray matter volume measurements in patients with AIDS based on magnetic resonance.The top ten brain regions in the AIDS patient group and the control group are mainly concentrated in the bilateral frontal lobe,bilateral parietal lobe,and left temporal lobe.The gray matter volume index can be used to diagnose HAND.The examination of magnetic resonance structure imaging can provide an objective examination method for clinical diagnosis of HAND.The examination of magnetic resonance structure imaging can provide objective evaluation indicators for the treatment effect,so as to reduce the occurrence and development of HAND.PART THREE: THE APPLICATION OF MACHINE LEARNING METHODTO EVALUATE THE DEGREE OF CENTRALITY AND VOXEL-MIRROR HOMOTOPIC CONNECTIVITY OF RESTING STATE FUNCTIONAL MAGNETIC RESONANCE IMAGING ABOUT NETWORK CONNECTIVITY OF AIDS PATIENTS.OBJECTIVE: To investigate the diagnostic value of DC values and VMHC values in the assessment of HIV-related neurocognitive disorders,and the DC values and VMHC values ? in each brain region are related to clinical indicators,respectively,using magnetic resonance resting function imaging based on machine learning.MATERIALS AND METHODS: The inclusion criteria for inclusion in the study were the same as in the first part.All subjects were right-handed and signed an informed consent form for magnetic resonance imaging using a GE Discovery MR 750 w 3.0T magnetic resonance scanner.The routine axial T1 WI scan,axial T2 WI scan,resting state functional magnetic resonance(rs-fMRI)scan,and sagittal T1 weighted(Sag3D T1WI-BRAVO)thin layer scan were performed.The image processing method is the same as the second part.Statistical processing is the same as the second part.RESULTS:(1)The ten brain regions with the greatest contribution to the difference in DC between AIDS patients and control groups in machine learning were the right paracentral lobule,the left paracentral lobule,the right superior occipital gyrus,the right supplementary motor area,the right cuneus,the left cuneus,the left superior parietal gyrus,the left precuneus,the left supplementary motor area,and the right postcentral gyrus.(2)The ten brain regions with the greatest contribution to the difference in VMHC between AIDS patients and control groups in machine learning were the left paracentral lobule,the right paracentral lobule,the right supplementary motor area,the right precuneus,theleft precuneus,the left inferior parietal gyrus,he left supplementary motor area,the left caudate nucleus,the left postcentral lobule,the left heschl gyrus.(3)The evaluation of the classification effect of DC and VMHC indicators was as follows: the ROC value of the DC index was 0.60,the accuracy rate was 65.85%,the sensitivity was 100%,the specificity was 0%,the positive predictive value was 65.85%,and the negative predictive value was 0%.The VMC indicator had a ROC value of 0.59,an accuracy rate of 68.85%,a sensitivity of 92.59%,a specificity of 14.29%,a positive predictive value of 67.57%,and a negative predictive value of 50.00%.(4)The DC index was meaningful for distinguishing between HAND group and non-HAND group.VMHC index was meaningless for distinguishing between HAND group and non-HAND group.(5)The DC values and VMHC values ? of the top ten brain regions were partially correlated with clinical hematology,clinical scales,cognitive scores,and groupings.CONCLUSION: The machine learning method shows that the DC index is more effective than the VMHC index,indicating that the application of DC index in AIDS patients is more meaningful.The DC values obtained by resting-state functional magnetic resonance imaging based on machine learning methods are meaningful for distinguishing between HAND and non-HAND partial brain regions,while the VMHC values of brain synchrony on both sides of the response are meaningless in distinguishing between HAND and non-HAND.This indicates that HIV-related cognitive.impairment is related to some brain network connections,and HIV-related cognitive impairment has little to do with the pathogenesis regulation mechanism of bilateral brain synchronization.The DC values and VMHC values of the top ten brain regions were correlated with clinical indicators(including hematology,clinical scales)and cognitive groupings.The DC value VMHC obtained by machine learning can directly reflect the extent of cognitive impairment and the corresponding functional damage,improve the clinical antiviral treatment plan and increase the intervention of neurocognitive disorders.The occurrence and development of HAND played a certain role.