ADNI Database Research Articles

Impact of serum leptin and adiponectin levels on the development of brain infarcts in patients with mild cognitive impairment and Alzheimer’s disease

AbstractBackgroundThe adipokines leptin and adiponectin have been associated with the development of atherosclerosis and the risk of cerebral infarctions. According to pre‐clinical studies, however, they play a protective role against ischemic brain damage. In this study we analysed the relationship between serum leptin and adiponectin levels and the onset or progression of silent brain infarctions in a population of subjects with mild cognitive impairment (MCI) and Alzheimer’s disease (AD).MethodAll data were extracted from the ADNI database. All patients with available serum leptin and adiponectin levels at baseline were selected. Demographic and anamnestic data, neuropsychological test results, tau, phosphotau and amyloid levels in CSF and regional brain metabolism with FDG‐PET data were also extracted. The determination of the number of silent cerebral infarctions was performed on baseline brain MRI. The final population includes 566 subjects, with 58 normal controls, 396 MCI and 112 AD.ResultLeptin levels were significantly lower in patients with MCI than controls at baseline, while adiponectin levels were not different between the three groups. Multivariate regression analysis in patients with cognitive impairment (MCI + AD, 508 subjects) showed that leptin levels were significantly associated with BMI, arterial hypertension and female sex. By stratifying the three diagnostic groups by the presence or absence of baseline infarctions, leptin levels were significantly higher in patients with AD and brain infarcts than in patients with AD without brain infarcts. However, multivariate logistic regression analysis at baseline for the presence or absence of brain infarcts did not confirm the predictive value of leptin. Multivariate longitudinal analysis did not attribute predictive value to serum leptin levels on the development of silent brain infarcts in this MCI and AD population.ConclusionThe evidence on the pathogenetic or protective role of adipokines on ischemic brain damage in the literature is mixed. In this population affected by MCI and AD, serum leptin and adiponectin levels are not associated with the development of brain infarctions; therefore, these results do not support the use of adipokines as biomarkers of cerebrovascular pathology in this population.

Differences in dynamic functional connectivity across the Alzheimer’s disease spectrum

AbstractBackgroundTo examine differences in dynamic functional connectivity (DFC) in cognitively normal subjects (CN), and patients with mild cognitive impairment (MCI) or Alzheimer’s disease (AD).MethodResting‐state functional magnetic resonance imaging (rsfMRI data of 76 participants (40 CN, 24 MCI, 11 AD) from the ADNI database (http://adni.loni.usc.edu/) was used for analysis. Groups were matched according to age, sex and education. rsfMRI data were pre‐processed using the CONN toolbox. For data reduction purposes, spatial independent and principal component analyses were performed on the normalized rsfMRI using GIFT. Resulting meaningful independent components were clustered into 13 functional resting‐state networks. Using GIFT, a sliding window approach was employed to determine dynamic functional connectivity states across the scan period of 9 minutes. Afterwards, non‐parametric t‐tests were performed comparing the groups in measures of DFC such as dwell time (i.e. staying in a state), number of transitions (i.e. switching between states), fraction time (i.e. total time spent in a state), and meta state characteristics.ResultThe DFC analysis resulted in 4 distinct functional connectivity states, which were characterized by either global connectivity patterns or distinct connectivity patterns between certain rs‐networks. Group comparisons yielded that one state was particularly more occupied by the AD group compared to the CN group (p = .02) and the MCI group (p = .001). The AD group showed a lower meta state span and number of meta states compares to the MCI and CN groups.ConclusionNetwork reorganization may influence dynamic functional connectivity patterns in pre‐ and clinical AD.

Risk factors of brain atrophy in cognitively normal elderly

AbstractBackgroundLifestyle‐related risk factors (e.g., smoking) and common health problems (e.g., hypertension) have been found to be associated with brain atrophy. However, it is still not clear about their independent contribution to brain atrophy, especially among the elderly with normal cognition. Here, we aim to identify the independent risk factors of brain atrophy in cognitively normal elderly.MethodWe included 419 elderly subjects with normal cognition (Table 1) from the ADNI database (ADNI1/ADNI2/ADNI‐GO) who had available data of lifestyle‐related risk factors (alcohol abuse, smoking) and common health problems (obesity, hypertension, high cholesterol, high triglycerides, sleep disturbance) and underwent 3D T1‐weighted MRI scans at baseline. The T1‐weighted images were processed with AccuBrain IV2.0 to quantify the brain volumetric measures as listed in Table 2. Univariate analyses were performed with ANCOVA to investigate the influence of the risk factors for each brain volumetric measure, while age, gender, education and APOE4 carrier were covaried out. The independent contribution of the risk factors were further identified by entering them together in a multivariate linear regression model as predictors of brain atrophy using backward elimination.ResultIn the univariate analyses, obesity showed the most widely influence of brain atrophy throughout the brain, followed by hypertension which contributed to atrophy in cerebellum and part of cortical gray matter (GM), alcohol abuse that contributed to atrophy in part of basal ganglia structures, high triglycerides for atrophy of pons and cerebellum, and smoking for medial temporal lobe atrophy (Table 3). When other risk factors were also considered (Table 4), the contribution of obesity to the atrophy of most brain regions remained significant. Hypertension, alcohol abuse and smoking also showed independent contributions (p<0.05) to the atrophy of several brain regions, while sleep disturbance (for brain parenchyma) and high triglycerides (for cerebellum) only presented mild independent contribution (0.05<p<0.10) to brain atrophy.ConclusionObesity presented most independent contribution to brain atrophy in cognitively normal elderly, followed by hypertension, alcohol abuse and smoking. These results may facilitate the interpretation of brain atrophy in normal subjects and intervention of reversible brain atrophy caused by specific risk factors.

Contributions of macrostructural and microstructural white matter damage to apathy in people with Alzheimer’s disease

AbstractBackgroundApathy is the most common neuropsychiatric symptom in Alzheimer’s disease (AD) that affects patients’ quality of life and prognosis. Previous functional neuroimaging studies have found multiple cerebral alterations suggesting a conceptualisation of apathy as a symptom resulting from a communication breakdown among functional neural networks involved in motivational‐affective processing. To test this hypothesis, this study investigated the association between white matter (WM) structural connectivity and apathy in AD.MethodSixty‐one patients with (AD‐AP) and 61 without apathy (AD‐NA) were identified from the ADNI database and matched for diagnosis, cognitive status, age and education. A group of 61 matched healthy older adults (HOA) was also included. Data on cognitive performance, cerebrospinal fluid (CSF) biomarkers, brain volumes and mean DTI indices of 50 WM tracts were extracted for all participants. Cognitive profiles and volumetric/DTI data were compared across groups using general linear models including sex as a covariate. Post hoc analyses were performed to investigate pair‐wise differences in all outcome measures.ResultBoth patient groups showed widespread cognitive deficits and neural alterations, lower amyloid beta and higher phosphorylated tau CSF levels compared to the HOA group. The direct AD‐AP vs AD‐NA comparison showed no significant differences in CSF biomarkers and MRI outcomes, but the AD‐AP group had more severe neuropsychiatric symptoms, lower scores on the Clock Test – copy and relied more on cues in the Logical Memory test – delayed recall. When the patient groups were compared to the HOA independently, the AD‐AP group had more extensive WM damage: 1) reduced WM and larger WM hyperintensity volume; 2) increased radial diffusivity and reduced fractional anisotropy primarily in the cingulum, superior longitudinal and inferior fronto‐occipital fasciculi and genu of the corpus callosum.ConclusionAD‐AP patients had WM damage mildly more severe than that observed in AD‐NA, especially in associative tracts in the frontal lobes and the cingulum. Such disruption in structural connectivity might affect crucial functional inter‐network communication resulting in motivational deficits and worse cognitive decline. These findings warrant further investigations to ascertain the relationship between WM damage and disruption of functional brain network dynamic interactions in AD patients with apathy.

Fused multi‐modal similarity network as prior in guiding brain imaging genetic association

AbstractBackgroundBrain imaging genetics aims to explore the genetic architecture underlying brain structure and functions. Recent studies showed that incorporation of prior knowledge, such as subject diagnosis information and brain regional correlation, can help identify significantly stronger imaging genetic associations. However, sometimes such information may be incomplete or even unavailable. In this project, we explore a new data‐driven prior knowledge that captures the subject‐level similarity matrix derived from fusion of multi‐modal similarity networks. It was built into the sparse canonical correlation analysis model, which is expected to identify brain imaging and genetic markers that contribute to the fused similarity matrix.MethodThe Amyloid and Tau imaging data were downloaded from the ADNI database. We have 800 and 291 subjects for amyloid and tau imaging respectively (Table. 1). SNPs with p<5×10−6 (n = 1080) were extracted from the large‐scale GWAS summary statistics from the IGAP. We built similarity networks for imaging and genetic data individually using normalized mutual information and then fused these network following (Fig. 1). Fused network was then applied as a prior knowledge in the discriminative sparse canonical correlation analysis (DSCCA) model. We also tested other prior knowledge in SCCA model for comparison including diagnosis group network and transcriptomic brain network. All methods were evaluated using ADNI genotype, amyloid and tau imaging data, where nested 5‐fold cross validation with the same set of training and test data was applied for fair comparison.ResultSCCA with fused network and diagnosis group showed better performances over others in both training and testing data set (Table. 2,3). In Fig. 2 are top 10 ROIs identified by SCCA with fused network, whose amyloid deposition are strongly associated with APOE SNP (rs429358). For tau, we identified left and right amygdala with tau accumulation strongly and consistently associated with rs429358 (fig.3).ConclusionAs prior knowledge incorporated into SCCA algorithm, fused similarity network can help improve the performance of imaging genetic association over traditional SCCA and that guided by the brain co‐expression network. It showed comparable performance as diagnosis information and thus could be a potential prior knowledge to consider when lack of diagnosis information.

FLAIR MRI biomarkers of the Normal‐Appearing Brain Matter (NABM) are related to APOE‐4 status and CSF markers Aβ42 and Tau

AbstractBackgroundReduced amyloid‐beta‐42 (Aβ42) and elevated total tau (t‐tau) protein concentrations in the CSF are characteristic of Alzheimer’s disease (AD) pathology [1]. Aβ42 contributes to extracellular plaque formation between neurons [2]. The total level of tau, a microtubule‐stabilizer in neuron axons [3], is associated with neuronal death [1]. Additionally, presence of ApoE4 (ε4) alleles is an established identifier for increased risk of AD [4]. Non‐invasive imaging‐based correlates of the CSF markers are lacking. This study examines correlations of novel FLAIR texture, intensity, and volume biomarkers, along with white matter lesion (WML) burden, with Aβ42, t‐tau, and ε4 status.MethodFLAIR volumes (n = 4044) from the ADNI database were used for analysis with CSF biomarkers. A subset (n = 921) was used for ε4 analysis based on available data. Normal‐appearing brain matter (NABM) was obtained by removing CSF and WML. Macrostructural damage (MAD), microstructural damage (MID), and integrity (MII) texture biomarkers [5] were extracted from each NABM volume (Fig. 1), along with median intensity, NABM/ICV ratio, and normalized WML volume [5]. Relationships between FLAIR biomarkers and AD markers were described using parametric statistical analyses. Statistical significance was set at p<0.05.ResultAll FLAIR biomarkers were statistically significantly correlated with CSF Aβ42 levels. All features except MID and WML burden were significantly correlated with CSF t‐tau levels (Table 1). MAD, MID, and WML burden showed negative correlations with Aβ42, while MAD showed a positive correlation with t‐tau. Opposite trends were seen with MII, intensity, and NABM/ICV (Fig. 2). Significant differences between groups without and with one ε4 allele were found in all FLAIR biomarkers except WML burden. Significant differences in MAD, MII and NABM/ICV were found between groups with one and two ε4 alleles (Table 2).ConclusionFLAIR biomarkers were inversely associated with CSF Aβ42 concentrations, suggesting they may be surrogate markers of Aβ plaque formation. They were associated with CSF t‐tau concentrations, suggesting they could also be indicators of GM degeneration and neuronal death. They could also be indicators for patients at increased risk of AD, as they demonstrated significant differences with the presence of ε4. Further longitudinal studies are required to establish the exact mechanism of neurodegeneration.

Improved SUVR calculation for [18F]‐AV45 amyloid pet imaging using a novel reference region approach

AbstractBackgroundUtilization of a data‐driven approach to find a more sensitive reference region (RR) for [18F]‐AV45 PET imaging that differentiates the spectrum of Alzheimer’s disease (AD) and improves especially longitudinal study designs.MethodData of 283 participants (135 amyloid‐negative cognitively normal (CN), and 148 amyloid‐positive AD) from the ADNI database (http://adni.loni.usc.edu/) was used. All [18F]‐AV45 scans were co‐registered, normalized, and skull‐stripped. The dataset was split into a training‐and test‐dataset. Voxel‐wise group comparisons were performed in the training‐set (75 CNs, 77 ADs) to identify a RR that is void off on‐target tracer uptake. Potential clusters were used to extract mean global SUVRs in the test‐dataset (60 CNs, 70 ADs). Effect sizes between novel and commonly used RRs were compared. Baseline and follow‐up data of 19 CNs, 36 participants with mild cognitive impairment, and 26 ADs was used to test whether the newly identified RR is more sensitive to assess longitudinal change than common RRs. Effect sizes of change in SUVR between baseline and follow‐up were used as metric of sensitivity.ResultThe training dataset yielded two novel RRs in the brainstem and the cerebellar white matter. These new RRs showed higher effect sizes compared with commonly used RRs. However, no significant differences in the effect sizes were observed when examining longitudinal change in SUVR compared with commonly used RRs.ConclusionThe data‐driven approach did not improve SUVR measurements for [18]‐AV45 imaging. However, our analyses proved that the commonly used RRs are also longitudinally stable and sensitive to differentiate clinical from non‐clinical groups.

Explainable Deep-Learning-Based Diagnosis of Alzheimer’s Disease Using Multimodal Input Fusion of PET and MRI Images

PurposeAlzheimer’s disease (AD) is a progressive, incurable human brain illness that impairs reasoning and retention as well as recall. Detecting AD in its preliminary stages before clinical manifestations is crucial for timely treatment. Magnetic Resonance Imaging (MRI) provides valuable insights into brain abnormalities by measuring the decrease in brain volume expressly in the mesial temporal cortex and other regions of the brain, while Positron Emission Tomography (PET) measures the decrease of glucose concentration in the temporoparietal association cortex. When these data are combined, the performance of AD diagnostic methods could be improved. However, these data are heterogeneous and there is a need for an effective model that will harness the information from both data for the accurate prediction of AD.MethodsTo this end, we present a novel heuristic early feature fusion framework that performs the concatenation of PET and MRI images, while a modified Resnet18 deep learning architecture is trained simultaneously on the two datasets. The innovative 3-in-channel approach is used to learn the most descriptive features of fused PET and MRI images for effective binary classification of AD.ResultsThe experimental results show that the proposed model achieved a classification accuracy of 73.90% on the ADNI database. Then, we provide an Explainable Artificial Intelligence (XAI) model, allowing us to explain the results.ConclusionOur proposed model could learn latent representations of multimodal data even in the presence of heterogeneity data; hence, the proposed model partially solved the issue with the heterogeneity of the MRI and PET data.

Higher performance for women than men in MRI-based Alzheimer’s disease detection

IntroductionAlthough machine learning classifiers have been frequently used to detect Alzheimer’s disease (AD) based on structural brain MRI data, potential bias with respect to sex and age has not yet been addressed. Here, we examine a state-of-the-art AD classifier for potential sex and age bias even in the case of balanced training data.MethodsBased on an age- and sex-balanced cohort of 432 subjects (306 healthy controls, 126 subjects with AD) extracted from the ADNI data base, we trained a convolutional neural network to detect AD in MRI brain scans and performed ten different random training-validation-test splits to increase robustness of the results. Classifier decisions for single subjects were explained using layer-wise relevance propagation.ResultsThe classifier performed significantly better for women (balanced accuracy 87.58pm 1.14%) than for men (79.05pm 1.27%). No significant differences were found in clinical AD scores, ruling out a disparity in disease severity as a cause for the performance difference. Analysis of the explanations revealed a larger variance in regional brain areas for male subjects compared to female subjects.DiscussionThe identified sex differences cannot be attributed to an imbalanced training dataset and therefore point to the importance of examining and reporting classifier performance across population subgroups to increase transparency and algorithmic fairness. Collecting more data especially among underrepresented subgroups and balancing the dataset are important but do not always guarantee a fair outcome.

Application of Nobel QSM technique in MRI for diagnosis of Alzheimer's disease: What is the relationship between iron deposits in brain nuclei with age and severity of disorders?

Background. Iron deposits impair neurons and cause cognitive disorders through different pathways. Today, with the help of quantitative susceptibility mapping (QSM), iron accumulations in different brain areas can be evaluated and measured.This research aims to identify the changes of iron deposits in the brain nuclei in different stages of cognitive disorders using the QSM technique and introduces biomarkers for the early diagnosis of these disorders. Methods. The data of thirty-five participants with normal cognitive conditions and forty six patients with cognitive disorders, classified into four different groups based on the severity of the disorder, were downloaded from the ADNI database. After determining the regions (ROIs) with the FSL automatic segmentation tool, the mean QSM values of twelve brain nuclei were determined with the help of Slicer software, and SPSS software was used for statistical analysis. Results. There was an inverse correlation in the left putamen nucleus of LMCI group (P=0.041). Direct correlation between age variable and QSM values of the left putamen nucleus (P=0.000078) and right putamen nucleus (P=0.000017) of the control group were reported. Conclusion. As the severity of disorders increases in the LMCI group, the number of paramagnetic deposits, such as iron, in the left putamen nucleus increases. Therefore, the QSM values of the left putamen nucleus can be used as an indicator of the severity of the disorder in the advanced stages of cognitive disorders. Practical Implications. Iron deposits cause damage to neurons and cause cognitive disorders through various pathways. Today, with the help of the quantitative susceptibility mapping technique, the numeral of iron accumulations in different brain areas can be evaluated and measured. This research aims to identify the changes of iron deposits in the brain nuclei in different stages of cognitive disorders by using the Quantitative susceptibility mapping (QSM) technique to introduce biomarkers for early diagnosis of these disorders.

A Comparative Study of Correlation Methods in Functional Connectivity Analysis Using fMRI Data of Alzheimer's Patients.

Functional Magnetic Resonance Imaging (fMRI) is a non-invasive neuroimaging tool, used in brain function research and is also a low-frequency signal, showing brain activation by means of Oxygen consumption. One of the reliable methods in brain functional connectivity analysis is the correlation method. In correlation analysis, the relationship between two time-series has been investigated. In fMRI analysis, the Pearson correlation is used while there are other methods. This study aims to investigate the different correlation methods in functional connectivity analysis. In this analytical research, based on fMRI signals of Alzheimer's Disease (AD) and healthy individuals from the ADNI database, brain functional networks were generated using correlation techniques, including Pearson, Kendall, and Spearman. Then, the global and nodal measures were calculated in the whole brain and in the most important resting-state network called Default Mode Network (DMN). The statistical analysis was performed using non-parametric permutation test. Results show that although in nodal analysis, the performance of correlation methods was almost similar, in global features, the Spearman and Kendall were better in distinguishing AD subjects. Note that, nodal analysis reveals that the functional connectivity of the posterior areas in the brain was more damaged because of AD in comparison to frontal areas. Moreover, the functional connectivity of the dominant hemisphere was disrupted more. Although the Pearson method has limitations in capturing non-linear relationships, it is the most prevalent method. To have a comprehensive analysis, investigating non-linear methods such as distance correlation is recommended.

Broad learning for early diagnosis of Alzheimer's disease using FDG-PET of the brain.

Alzheimer's disease (AD) is a progressive neurodegenerative disease, and the development of AD is irreversible. However, preventive measures in the presymptomatic stage of AD can effectively slow down deterioration. Fluorodeoxyglucose positron emission tomography (FDG-PET) can detect the metabolism of glucose in patients' brains, which can help to identify changes related to AD before brain damage occurs. Machine learning is useful for early diagnosis of patients with AD using FDG-PET, but it requires a sufficiently large dataset, and it is easy for overfitting to occur in small datasets. Previous studies using machine learning for early diagnosis with FDG-PET have either involved the extraction of elaborately handcrafted features or validation on a small dataset, and few studies have explored the refined classification of early mild cognitive impairment (EMCI) and late mild cognitive impairment (LMCI). This article presents a broad network-based model for early diagnosis of AD (BLADNet) through PET imaging of the brain; this method employs a novel broad neural network to enhance the features of FDG-PET extracted via 2D CNN. BLADNet can search for information over a broad space through the addition of new BLS blocks without retraining of the whole network, thus improving the accuracy of AD classification. Experiments conducted on a dataset containing 2,298 FDG-PET images of 1,045 subjects from the ADNI database demonstrate that our methods are superior to those used in previous studies on early diagnosis of AD with FDG-PET. In particular, our methods achieved state-of-the-art results in EMCI and LMCI classification with FDG-PET.

Data Augmentation in High Dimensional Low Sample Size Setting Using a Geometry-Based Variational Autoencoder.

In this paper, we propose a new method to perform data augmentation in a reliable way in the High Dimensional Low Sample Size (HDLSS) setting using a geometry-based variational autoencoder (VAE). Our approach combines the proposal of 1) a new VAE model, the latent space of which is modeled as a Riemannian manifold and which combines both Riemannian metric learning and normalizing flows and 2) a new generation scheme which produces more meaningful samples especially in the context of small data sets. The method is tested through a wide experimental study where its robustness to data sets, classifiers and training samples size is stressed. It is also validated on a medical imaging classification task on the challenging ADNI database where a small number of 3D brain magnetic resonance images (MRIs) are considered and augmented using the proposed VAE framework. In each case, the proposed method allows for a significant and reliable gain in the classification metrics. For instance, balanced accuracy jumps from 66.3% to 74.3% for a state-of-the-art convolutional neural network classifier trained with 50 MRIs of cognitively normal (CN) and 50 Alzheimer disease (AD) patients and from 77.7% to 86.3% when trained with 243 CN and 210 AD while improving greatly sensitivity and specificity metrics.

Based on Tau PET Radiomics Analysis for the Classification of Alzheimer’s Disease and Mild Cognitive Impairment

Alzheimer's Disease (AD) and Mild Cognitive Impairment (MCI) are closely associated with Tau proteins accumulation. In this study, we aimed to implement radiomics analysis to discover high-order features from pathological biomarker and improve the classification accuracy based on Tau PET images. Two cross-racial independent cohorts from the ADNI database (121 AD patients, 197 MCI patients and 211 normal control (NC) subjects) and Huashan hospital (44 AD patients, 33 MCI patients and 36 NC subjects) were enrolled. The radiomics features of Tau PET imaging of AD related brain regions were computed for classification using a support vector machine (SVM) model. The radiomics model was trained and validated in the ADNI cohort and tested in the Huashan hospital cohort. The standard uptake value ratio (SUVR) and clinical scores model were also performed to compared with radiomics analysis. Additionally, we explored the possibility of using Tau PET radiomics features as a good biomarker to make binary identification of Tau-negative MCI versus Tau-positive MCI or apolipoprotein E (ApoE) ε4 carrier versus ApoE ε4 non-carrier. We found that the radiomics model demonstrated best classification performance in differentiating AD/MCI patients and NC in comparison to SUVR and clinical scores models, with an accuracy of 84.8 ± 4.5%, 73.1 ± 3.6% in the ANDI cohort. Moreover, the radiomics model also demonstrated greater performance in diagnosing AD than other methods in the Huashan hospital cohort, with an accuracy of 81.9 ± 6.1%. In addition, the radiomics model also showed the satisfactory classification performance in the MCI-tau subgroup experiment (72.3 ± 3.5%, 71.9 ± 3.6% and 63.7 ± 5.9%) and in the MCI-ApoE subgroup experiment (73.5 ± 4.3%, 70.1 ± 3.9% and 62.5 ± 5.4%). In conclusion, our study showed that based on Tau PET radiomics analysis has the potential to guide and facilitate clinical diagnosis, further providing evidence for identifying the risk factors in MCI patients.

AlzheimerNet: An Effective Deep Learning Based Proposition for Alzheimer’s Disease Stages Classification From Functional Brain Changes in Magnetic Resonance Images

Alzheimer’s disease is largely the underlying cause of dementia due to its progressive neurodegenerative nature among the elderly. The disease can be divided into five stages: Subjective Memory Concern (SMC), Mild Cognitive Impairment (MCI), Early MCI (EMCI), Late MCI (LMCI), and Alzheimer’s Disease (AD). Alzheimer’s disease is conventionally diagnosed using an MRI scan of the brain. In this research, we propose a fine-tuned convolutional neural network (CNN) classifier called AlzheimerNet, which can identify all five stages of Alzheimer’s disease and the Normal Control (NC) class. The ADNI database’s MRI scan dataset is obtained for use in training and testing the proposed model. To prepare the raw data for analysis, we applied the CLAHE image enhancement method. Data augmentation was used to remedy the unbalanced nature of the dataset and the resultant dataset consisted of 60000 image data on the 6 classes. Initially, five existing models including VGG16, MobileNetV2, AlexNet, ResNet50 and InceptionV3 were trained and tested to achieve test accuracies of 78.84%, 86.85%, 78.87%, 80.98% and 96.31% respectively. Since InceptionV3 provides the highest accuracy, this model is later modified to design the AlzheimerNet using RMSprop optimizer and learning rate 0.00001 to achieve the highest test accuracy of 98.67%. The five pre-trained models and the proposed fine-tuned model were compared in terms of various performance matrices to demonstrate whether the AlzheimerNet model is in fact performing better in classifying and detecting the six classes. An ablation study shows the hyperparameters used in the experiment. The suggested model outperforms the traditional methods for classifying Alzheimer’s disease stages from brain MRI, as measured by a two-tailed Wilcoxon signed-rank test, with a significance of < 0.05.

Dynamic changes of CSF clusterin levels across the Alzheimer’s disease continuum

BackgroundClusterin is a multifunctional protein, which is associated with the pathogenesis and the development of Alzheimer’s disease (AD). Compared with normal controls, inconsistent results have yielded in previous studies for concentration of cerebrospinal fluid (CSF) clusterin in AD patients. We explored CSF clusterin levels in different pathological processes of AD.MethodsFollowing the National Institute on Aging-Alzheimer’s Association (NIA-AA) criteria, we employed on the levels of CSF Aβ42(A), phosphorylated-Tau (T), and total-tau (N). Based on previously published cutoffs and the close correlation between CSF p-tau and t-tau, 276 participants from the publicly available ADNI database with CSF biomarkers were divided into four groups: A-(TN)- (normal Aβ42 and normal p-tau and t-tau; n = 50), A+(TN)- (abnormal Aβ42 and normal p-tau and t-tau; n = 39), A+(TN) + (abnormal Aβ42 and abnormal p-tau or t-tau; n = 147), A-(TN) + (normal Aβ42 and abnormal p-tau or t-tau; n = 40). To assess CSF clusterin levels in AD continuum, intergroup differences in four groups were compared. Pairwise comparisons were conducted as appropriate followed by Bonferroni post hoc analyses. To further study the relationships between CSF clusterin levels and AD core pathological biomarkers, we employed multiple linear regression method in subgroups.ResultsCompared with the A-(TN)- group, CSF clusterin levels were decreased in A+ (TN)- group (P = 0.002 after Bonferroni correction), but increased in the A+(TN) + group and the A-(TN) + group (both P < 0.001 after Bonferroni correction). Moreover, we found CSF clusterin levels are positively associated with CSF Aβ42 (β = 0.040, P < 0. 001), CSF p-tau (β = 0.325, P < 0.001) and CSF t-tau (β = 0.346, P < 0.001).ConclusionsOur results indicated that there are differences levels of CSF clusterin in different stages of AD pathology. The CSF clusterin level decreased at the early stage are related to abnormal Aβ pathology; and the increased levels are associated with tau pathology and neurodegeneration.

Soluble TREM2 levels associate with conversion from mild cognitive impairment to Alzheimer’s disease

BACKGROUNDSoluble triggering receptor expressed on myeloid cells 2 (sTREM2) plays an important role in the clearance of pathological amyloid-β (Aβ) in Alzheimer's disease (AD). This study aimed to explore sTREM2 as a central and peripheral predictor of the conversion from mild cognitive impairment (MCI) to AD.METHODSsTREM2 and Aβ1-42 levels in cerebrospinal fluid (CSF) and florbetapir-PET (AV45) images were analyzed for healthy control (HCs), patients with MCI, and patients with AD from the ADNI database. Peripheral plasma sTREM2 and Aβ1-42 levels were determined for our Neurology database of Ruijin Hospital for Alzheimer's Disease (NRHAD) cohort, and patients with MCI were reevaluated at follow-up visits to assess for progression to AD. The association between CSF and plasma sTREM2 levels was analyzed in data from the Chinese Alzheimer's Biomarker and Lifestyle (CABLE) database.RESULTSThe results showed that patients with MCI who had low levels of CSF sTREM2 and Aβ1-42 were more likely to develop AD. Among participants with positive Aβ deposition, as assessed by AV45 imaging, elevated CSF sTREM2 levels were associated with a decreased risk of MCI-to-AD conversion. Meanwhile, in the NRHAD cohort, individuals in the MCI group with high sTREM2 levels in plasma were at a greater risk for AD, whereas low Aβ1-42 with high sTREM2 levels in plasma were associated with a faster cognitive decline. In addition, CSF sTREM2 levels were highly correlated with plasma sTREM2 levels in the CABLE database.CONCLUSIONThese findings suggest that sTREM2 may be useful as a potential predictive biomarker of MCI-to-AD conversion.FUNDINGThis study was supported by grants from the National Natural Science Foundation of China (grant nos. 82001341, 82071415, 81873778, and 82201392); the Shanghai Sailing Program (grant no. 22YF1425100); and the China Postdoctoral Science Foundation funded project (grant no. 2021M702169).

Multi-band network fusion for Alzheimer's disease identification with functional MRI.

The analysis of functional brain networks (FBNs) has become a promising and powerful tool for auxiliary diagnosis of brain diseases, such as Alzheimer's disease (AD) and its prodromal stage. Previous studies usually estimate FBNs using full band Blood Oxygen Level Dependent (BOLD) signal. However, a single band is not sufficient to capture the diagnostic and prognostic information contained in multiple frequency bands. To address this issue, we propose a novel multi-band network fusion framework (MBNF) to combine the various information (e.g., the diversification of structural features) of multi-band FBNs. We first decompose the BOLD signal adaptively into two frequency bands named high-frequency band and low-frequency band by the ensemble empirical mode decomposition (EEMD). Then the similarity network fusion (SNF) is performed to blend two networks constructed by two frequency bands together into a multi-band fusion network. In addition, we extract the features of the fused network towards a better classification performance. To verify the validity of the scheme, we conduct our MBNF method on the public ADNI database for identifying subjects with AD/MCI from normal controls. Experimental results demonstrate that the proposed scheme extracts rich multi-band network features and biomarker information, and also achieves better classification accuracy.

Hearing impairment is associated with cognitive decline, brain atrophy and tau pathology.

Hearing impairment was recently identified as the most prominent risk factor for dementia. However, the mechanisms underlying the link between hearing impairment and dementia are still unclear. We investigated the association of hearing performance with cognitive function, brain structure and cerebrospinal fluid (CSF) proteins in cross-sectional, longitudinal, mediation and genetic association analyses across the UK Biobank (N=165,550), the Chinese Alzheimer's Biomarker and Lifestyle (CABLE, N=863) study, and the Alzheimer's Disease Neuroimaging Initiative (ADNI, N=1770) database. Poor hearing performance was associated with worse cognitive function in the UK Biobank and in the CABLE study. Hearing impairment was significantly related to lower volume of temporal cortex, hippocampus, inferior parietal lobe, precuneus, etc., and to lower integrity of white matter (WM) tracts. Furthermore, a higher polygenic risk score (PRS) for hearing impairment was strongly associated with lower cognitive function, lower volume of gray matter, and lower integrity of WM tracts. Moreover, hearing impairment was correlated with a high level of CSF tau protein in the CABLE study and in the ADNI database. Finally, mediation analyses showed that brain atrophy and tau pathology partly mediated the association between hearing impairment and cognitive decline. Hearing impairment is associated with cognitive decline, brain atrophy and tau pathology, and hearing impairment may reflect the risk for cognitive decline and dementia as it is related to bran atrophy and tau accumulation in brain. However, it is necessary to assess the mechanism in future animal studies. A full list of funding bodies that supported this study can be found in the Acknowledgements section.

Brain morphometric and metabolic changes in Subjective Cognitive Decline individuals

AbstractBackgroundSubjective cognitive decline (SCD) is defined as a memory complaint in individuals without an objective measure of cognitive impairment. In this study, we investigated the morphometric and metabolic brain changes in individuals presenting with SCD.MethodAge and sex matched structural MRI and [18F]FDG PET images from SCD and CU (n = 101, per group) were extracted from the ADNI database. We built general linear models corrected for multiple comparisons with 10,000 Monte‐Carlo simulations to verify differences in cortical thickness between groups, using age as a covariate (Freesurfer’s v7.1.1). Metabolic analysis was performed using voxel‐wise general linear models comparing the SUVR of the FDG signal, correcting for age and APOE4 allele carrying status, using the MINC toolkit and the R statistical program.ResultSCD and CU groups presented similar age, sex, years of education, CSF AD‐related protein levels (amyloid beta 1‐42, p‐tau and tau), and Clinical Dementia Rating – Sum of Boxes score (Table 1). Compared to CU individuals, SCD subjects showed increased cortical thickness in the right hemisphere rostral middle frontal gyrus (Figure 1), while glucose metabolism was increased in the right medial temporal lobe, including the hippocampus, and right occipital cortex (Figure 2).ConclusionIndividuals with SCD present cortical hypertrophy in the rostral middle frontal gyrus and glucose hypermetabolism in the medial temporal lobe and occipital cortex, all in the right hemisphere. Previous studies suggest that these areas play a role in personality and humor, and may be altering self‐perception of cognitive decline. At the biological level, given the high concentration of astrocytes in these areas, such changes may suggest a scenario of astrocyte reactivity. More specifically, astrocyte hypertrophy may account for increased cortical thickness and glucose uptake. In summary, we provide neuroimaging evidence of structural and functional changes in individuals with SCD, suggesting early astrocyte reactivity.