Development And Clinical Application Of Prognostic Model For Progression From Amnestic Mild Cognitive Impairment To AD Based On Neuroimaging Characteristics

BackgroundMild cognitive impairment(MCI)has been conceptualized as a transitional clinical state between normal aging and AD-type dementia.Numerous studies have indicated that rates of MCI progress to AD are in the 10-15%per year range,and are 10 times than those of normal aging(1-2%per year).MCI can be further classified into amnestic MCI(aMCI)/non-amnestic MCI(naMCI).The aMCI subtypes,characterized as memory impairment,are considered to have a higher conversion rate to AD than other subtypes.Therefore,MCI has been considered as a key prognostic and therapeutic target in the management of AD.However,MCI has proved to be a highly heterogeneous syndrome with varying clinical outcomes.Not all MCI subjects convert to AD and many individuals remain cognitively stable or revert to normal status.For clinicians,early detection of those MCI individuals who are more likely to progress to probable AD is of increasing clinical importance in the enrichment of clinical trials of disease-modifying therapies.To date,although significant progress in building effective and accurate biomarker prognostic models that predict the conversion from MCI to AD has been made by adapting well-known machine learning tools,including support vector machine,logistic regression,boosting,and other pattern classification methods,some studies usually employed single modal biomarker and have ignored that the incorporation of multi-modal biomakers can add further predictive information to the prognostic model.More particularly,the conclusions were made base on the comparison of between-group with statistical significant level.The dichotomizing biomarker status based on the cut-off point threshold appear to be difficult to standardize across laboratories and may deviate from the continuous nature of insidious AD progression.Furthermore,although some studies have integrated multi-modal biomarkers to predict the MCI-to-AD conversion based on individual level,these studies have a limited prediction accuracy in differentiating individuals with MCI who converted to AD from non-converters.Cerebrospinal fluid(CSF)and positron emission tomography(PET)biomarkers are often not available in clinical practice since obtaining these biomarkers(e.g.CSF,PET,and so on)is laborious for the patient and clinician,and,at the same time,induces delays and increases the costs of the diagnosis significantly.Furthermore,CSF and PET sample collections are considered invasive.More especially,the performance of the prodictive model is not validated in an independent population for more precise estimation on how accurately they will generalize in practice.Therefore,converging evidence suggests prediction accuracies of progression from MCI to AD should be improved and there is a great need of sensitive,generalized,widely available,cost-effective,and minimally-invasive tools for the early detection of AD progression.Therefore,this study aimed to(1)explore and determine the susceptible resting-state neuroimaging characteristics(i.e.biomarker characteristics)that can predict the aMCI-to-AD conversion,as characterizing AD risk biomarker event;(2)in Alzheimer's Disease Neuroimaging Initiative(ADNI)dataset,integrate widely available,cost-effective,non-invasive AD risk markers from behavioral,brain,structural,and functional levels by using event-based probabilistic model and build the optimal temporal sequence of these biomarkers,then develop an highly accurate characterizing AD risk event index(CARE index)that is high sensitivity and specificity for predicting MCI individuals who will develop AD over 3 years;(3)apply the CARE index to distinguish those MCI individuals who progressed to AD-type dementia from those who did not from the independent Nanjing Aging and Dementia Study(NADS)during a 3-year follow-up period to explore the between-cohort generalization during the same time-period,further generate a precise cut-off value that separates MCI converters from MCI non-converters to guide the early diagnosis and disease-modifying treatments,then investigate the relationships between CAER index and cognitive performance or clinical variables in MCI patients to assess the behavioral significance of the CARE index,and finally promote the transformation of clinical use.Part 1:Rest-state neuroimaging characteristics of amnestic mild cognitive impairment at baselineChapter 1:The characteristics on brain structural atrophy in patients with amnestic mild cognitive impairmentObjective To assess the interaction of apolipoprotein E(APOE)by the aging process on brain morphology in aMCI patients.Methods Total 185 elderly subjects including 85 aMCI and 100 healthy controls(HC)were recruited.All subjects underwent a standardized clinical interview and comprehensive neuropsychological assessments,and APOE genotyping.We performed the analysis of gray matter(GM)voxel-based morphometry.Multivariate regression were used to investigate the interaction of APOE genotypes by age,and the relationships between brain atophy and cognitive performance.Results Compared with HC,aMCI patients showed multi-modal cognitive domain deficits,including episodic memory,information procesing speed,executive function,perceptual speed,working memory,and visuo-spatial cognition.Compared with HC,aMCI patients showed decreased GM volume in the left cerebellum anterior and cerebellum posterior lobe and the left hippocampus and parahippocampal gyrus.There was a interaction of group by APOE genotype in the left medial frontal gyrus,the left insula and the left calcarine.The multivariate regression analysis showed a negative correlation for APOE ?4-carriers and a positive correlation for APOE?2-carriers(except the left insula),while no correlations were found for APOE ?3/?3 between age and GM volumes on above brain regions.Moreover,the reduced GM volumes in the left calcarine and insula correlated with the impairment of visuo-spatial cognition and episodic memory in ?4-and ?2-carriers but not ?3/?3.respectively.Conclusions These results suggest that the APOE ?4 and ?2 alleles have the opposing effects on brain morphology across the spectrum of cognitive aging.Moreover,the interaction of APOE by age on brain morphology may accelerate the pathological progression of late-life cognitive decline in aMCI with c4-carriers and delay the possible conversion from aMCI with ?2-carriers to AD.Chapter 2:The characteristics on large-scale functional connectivity of subregions of medial temporal lobe in patients with amnestic mild cognitive impairmentObjective To investigate the functional circuitry of multiple subdivisions of parahippocampal gyrus(PHG)and hippocampus(HIP)and examine how this knowledge contributes to a more principled understanding of the contributions of its subregions to memory in aMCI.Methods We created two parallel arrays of four-millimeter spherical seeds along the longitudinal axis of the PHG and HIP respectively,followed by coregistered to the functional data.A total of 7 seeds were defined in the left hemisphere.Specifically,the four seeds located appropriately at the perirhinal cortex(i.e.PRC),the tranisition area of entorhinal cortex(ERC)and posterior PRC region(i.e.ERC/pPRC),the anterior and posterior parahippocampal cortex(i.e.aPHC,pPHC).three other seeds located appropriately at the head(anterior),body(middle)and tail(posterior)of the HIP(i.e.aHIP,mHIP,pHIP)respecitively.The large-scale functional connectivity(FC)analysis was performed in 85 aMCI and 129 healthy controls(HC).The aMCI demonstrated the distinctly disruptive patterns of the MTL subregional connectivity with the whole-brain.Results Compared with HC subjects,the aMCI patients demonstrated the asymmetric damage between left and right hemispheres and distinct disruptive patterns along the anterior through middle to posterior axis of the medial temporal lobe(MTL)subregional connectivity with the widely distributed cortical and subcortical regions,which suggests the impairment of the functional integration in the MTL.Notably,the right entorhinal cortex(ERC),middle HIP and perirhinal cortex(PRC)networks showed increased connectivity with the left occipital-temporal pathway,which potentially indicates a compensatory mechanism.Interestingly,the abnormal synchronicity of the damage degree in the MTL subregional networks was found in these regions of abnormal connectivity,namely there was a similar narrowing trend in the FC strength.Furthermore,the right abnormal MTL subregional FC changes were closely associated with cognitive performance in aMCI patients.The relationship between the mean increased FC in the whole-brain and the aging process present an inverse U-shaped curve with its peak locating at around 70 years old.Conclusions These results provide novel insights into the heterogeneous nature of its large-scale connectivity in MTL subregions in memory system underlying the memory deficits in aMCI.It further suggests that altered FC of MTL subregions are associated with the impairment of the differential encoding stages of memories and the functional changes in the specific right HIP-ERC-PRC-temporal circuitry may contribute to the impairment of episodic memory in aMCI.Chapter 3:The characteristics on intranetwork and internetwork connectivity patterns in patients with amnestic mild cognitive impairmentObjective Resting-state functional connectivity magnetic resonance imaging(rs-fcMRI)was performed to define the aberrant patterns in intranetwork and internetwork connectivity in aMCI and to examine how this knowledge contributes to a more essential understanding of the altered sequences involved in functional systems inside and outside of resting-state networks.Methods We investigated the functional connectivity patterns of five well-defined resting-state networks(RSNs)including default-mode network(DMN),dorsal attention network(DAN),control network(CON),salience network(SAL),sensory-motor network(SMN)from three levels(integrity level,network level,and edge level)in 87 aMCI and 114 healthy controls(HC),and to investigate the links between behavior and altered connectivity patterns at each level,partial correlation analyses were used to evaluate the correlations between the altered functional connectivity of each level and the clinical variables in aMCI.Results Compared with HC,aMCI showed only focal functional changes in regions of interest pairs,a trend toward increased correlations within the salience network and SMN,and a trend toward reduced correlation in the DMN-CON pair.Furthermore,these altered connectivities were associated with specific multi-domain cognitive and behavioral functions in aMCI.Notably,altered connectivity between right middle temporal cortex and posterior cerebellum was negatively correlated with Mattis Dementia Rating Scale scores in aMCI.Conclusions These results demonstrate that aMCl patients present aberrant intranetwork and internetwork connectivity patterns.It further suggests that dysfunction in right specific temporal-cerebellum neural circuit may contribute to the conversion from aMCI to AD.Part 2:Developing prognostic model for progression from amnestic mild cognitive impairment to ADChapter 4:Building the sequential biological stages of Alzheimer' s disease riskObjective To build the sequential biological stages of AD risk events from behavioral,brain structural and functional,CSF levels.Methods We selected 45 healthy normal(CN)and 25 AD from ANDI dataset,and integrated 10 AD risk events from behavioral,brain structural and functional,CSF levels by event-based probabilistic model,EBP).Then we assessed the optimal sequence of 10 AD risk events occurrence,which was defined as characterizing Alzheimer's disease risk events index(CARE index).Results In the optimal temporal sequence,these biomarkers occur in turn as follow:hippocampal(HIP)and posterior cingulated cortex(PCC)functional network,aberrant CSF ?-amyloid and p-tau levels,cognitive deficit(i.e.mini-mental state examination,MMSE,and 13-item Alzheimer's Disease Assessment Scale-Cognitive Subscale,ADAS),HIP gray matter loss,episodic memory deficit(i.e.Auditory Verbal Learning Test,AVLT),fusiform gyrus(FUS)gray matter and functional network abnormality.Conclusions This study suggest that the CARE index can measure AD risk stage based on individual level.Chapter 5:Developing a composite biomarker prognostic model predicting progression from amnestic mild cognitive impairment to Alzheimer's disease at the individual basisObjective To develop a composite biomarker prognostic model that can accurately predict MCI-to-AD conversion over 3-years at the individual patient level.Methods We selected 74 aMCI subjects from ADNI dataset.Using an event-based probabilistic model(EBP)approach to integrate behavioral,brain structural,and functional biomarkers,we developed the Characterizing Alzheimer's disease Risk Events(CARE)index to measure disease stage.Using the receiver operating characteristic(ROC)analysis,we then applied the CARE index to distinguish those MCI individuals who progressed to AD-type dementia from those who did not from the ADNI during a 3-year follow-up period.Results The CARE index achieved a high prediction performance with 80.4%accuracy,75%sensitivity,82%specificity,0.809 AUC on MCI subjects from the ADNI database.Furthermore,the CARE index showed considerably better performance compared to the case of using individual biomarkers.Thirdly,we generated a precise cut-off value of CARE index(cut-off threshold = 6.54)that can accurately predict MCI-to-AD conversion over 3 years.Conclusions We developed an accurate prognostic index(CARE index)that predicts aMCI individuals who will develop AD over 3 years by integrating widely available,cost-effective,non-invasive markers from behavioral,brain structural,and functional levels.It further suggests that the CARE index can be usefully applied in the selection of individuals with MCI for clinical trials and identify the future conversion of MCI to AD for early disease-modifying treatment.Part 3:Clinical application of prognostic model for progression from amnestic mild cognitive impairment to ADChapter 6:Predicting predicting progression from amnestic mild cognitive impairment to Alzheimer's disease at the individual basis using CARE indexObjective To investigate whether the CARE index can be generalized to an independent Nanjing Aging and Dementia Study(NADS)cohort.Methods We selected 87 aMCI subjects from NADS dataset.Using an EBP approach to integrate behavioral,brain structural,and functional biomarkers,we built the CARE index to measure disease stage.Using the ROC analysis,the CARE index was subsequently generalized to the prediction of MCI individuals from an independent NADS dataset during 3-year follow-up.Results The CARE index achieved a highly validated prediction performance with 87.5%accuracy,81%sensitivity,90%specificity,85.5%balanced accuracy,0.861 AUC on aMCI subjects from the NADS dataset.Most especially,the CARE index power to discriminate MCI converters from MCI non-converters generalized well on the independent test cohort and showed a high robustness across-datasets.Furthermore,the CARE index showed considerably better performance compared to the case of using individual biomarkers.The precise cut-off value of CARE index from ADNI dataset achieved high robustness and generalization across datasets.Finally,the CARE index was significantly associated with clinical cognitive decline.Conclusions The CARE index can predict the aMCl-to-AD conversion over 3-year follow-up and is applicable across datasets.This suggests that the CARE index can be usefully applied in the selection of individuals with MCI for clinical trials and identify the future conversion of MCI to AD for early disease-modifying treatment.