Article
To read the full-text of this research, you can request a copy directly from the authors.

Abstract

Despite subjects with Dominantly-Inherited Alzheimer's Disease (DIAD) represent less than 1% of all Alzheimer's Disease (AD) cases, the Dominantly Inherited Alzheimer Network (DIAN) initiative constitutes a strong impact in the understanding of AD disease course with special emphasis on the presyptomatic disease phase. Until now, the 3 genes involved in DIAD pathogenesis (PSEN1, PSEN2 and APP) have been commonly merged into one group (Mutation Carriers, MC) and studied using conventional statistical analysis. Comparisons between groups using null-hypothesis testing or longitudinal regression procedures, such as the linear-mixed-effects models, have been assessed in the extant literature. Within this context, the work presented here performs a comparison between different groups of subjects by considering the 3 genes, either jointly or separately, and using tools based on Machine Learning (ML). This involves a feature selection step which makes use of ANOVA followed by Principal Component Analysis (PCA) to determine which features would be realiable for further comparison purposes. Then, the selected predictors are classified using a Support-Vector-Machine (SVM) in a nested k-Fold cross-validation resulting in maximum classification rates of 72-74% using PiB PET features, specially when comparing asymptomatic Non-Carriers (NC) subjects with asymptomatic PSEN1 Mutation-Carriers (PSEN1-MC). Results obtained from these experiments led to the idea that PSEN1-MC might be considered as a mixture of two different subgroups including: a first group whose patterns were very close to NC subjects, and a second group much more different in terms of imaging patterns. Thus, using a k-Means clustering algorithm it was determined both subgroups and a new classification scenario was conducted to validate this process. The comparison between each subgroup vs. NC subjects resulted in classification rates around 80% underscoring the importance of considering DIAN as an heterogeneous entity.

No full-text available

Request Full-text Paper PDF

To read the full-text of this research,
you can request a copy directly from the authors.

... This experiment used the DIAN database (246 samples, 722 features), which was conceived for prediction of the evolution of DIAD, as well as to analyse the genetic subgroups stratified by specific mutations. For these purposes, the most commonly used techniques are based on machine learning algorithms [59,60], although studies using artificial neural networks have also been published [61]. ...
... It should be noted that the results associated with the classification of the DIAN dataset were not excessively high due to the complexity of the data, as most of the subjects were young people with hardly any symptoms of Alzheimer's disease. Nonetheless, the values obtained were similar to those achieved in previous studies [59]. RUB accuracies were slightly lower than using resubstitution due to the upper bound correction, and therefore more conservative than using CV (e.g. ...
... Firstly, only data from the initial visit of the subjects is considered, which reduces the number of samples from 1219 to 534 samples. Then, 29 patients were excluded as in [59], due to being diagnosed with at least one of the following diseases: cerebral stroke (3 subjects), transient ischemic attack (2 subject), dementia by alcoholism (4 subjects), Parkinson's disease (1 subject), traumatic brain injury with chronic deficit/dysfunction (3 subjects), dementia with Lewy bodies (1 subject), vascular dementia (1 subject) and dementia by unknown causes (5 subjects). Besides, in order not to increase the heterogeneity in symptomatic subjects, Late Onset AD (LOAD) cases in the DIAN study have been also discarded (15 subjects). ...
Preprint
Full-text available
Discriminative analysis in neuroimaging by means of deep/machine learning techniques is usually tested with validation techniques, whereas the associated statistical significance remains largely under-developed due to their computational complexity. In this work, a non-parametric framework is proposed that estimates the statistical significance of classifications using deep learning architectures. In particular, a combination of autoencoders (AE) and support vector machines (SVM) is applied to: (i) a one-condition, within-group designs often of normal controls (NC) and; (ii) a two-condition, between-group designs which contrast, for example, Alzheimer's disease (AD) patients with NC (the extension to multi-class analyses is also included). A random-effects inference based on a label permutation test is proposed in both studies using cross-validation (CV) and resubstitution with upper bound correction (RUB) as validation methods. This allows both false positives and classifier overfitting to be detected as well as estimating the statistical power of the test. Several experiments were carried out using the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset, the Dominantly Inherited Alzheimer Network (DIAN) dataset, and a MCI prediction dataset. We found in the permutation test that CV and RUB methods offer a false positive rate close to the significance level and an acceptable statistical power (although lower using cross-validation). A large separation between training and test accuracies using CV was observed, especially in one-condition designs. This implies a low generalization ability as the model fitted in training is not informative with respect to the test set. We propose as solution by applying RUB, whereby similar results are obtained to those of the CV test set, but considering the whole set and with a lower computational cost per iteration.
... FMRI modalities have been used to analyze brain connectivity patterns in diagnosing various brain disorders by physicians and have yielded promising results [26][27][28]. In recent years, studies have shown that functional connectivity analysis based on fMRI modalities plays a significant role in the diagnosis of brain disorders such as SZ [6], Alzheimer's disease (AD) [29], epilepsy [30][31], ADHD [5], and BD [32]. Brain disorders alter functional connectivity in the brain, which can be seen by fMRI modalities. ...
... The introduction of DL techniques in various fields, including medicine [25][26][27][28][29][30], has made significant progress, and valuable results have been achieved in detecting different diseases [33][34]. So far, several models of DL techniques have been proposed, with different training schemes, including supervised, semi-supervised, and unsupervised methods [54][55]. ...
Preprint
Full-text available
Nowadays, many people worldwide suffer from brain disorders, and their health is in danger. So far, numerous methods have been proposed for the diagnosis of Schizophrenia (SZ) and attention deficit hyperactivity disorder (ADHD), among which functional magnetic resonance imaging (fMRI) modalities are known as a popular method among physicians. This paper presents an SZ and ADHD intelligent detection method of resting-state fMRI (rs-fMRI) modality using a new deep learning (DL) method. The University of California Los Angeles (UCLA) dataset, which contains the rs-fMRI modalities of SZ and ADHD patients, has been used for experiments. The FMRIB software library (FSL) toolbox first performed preprocessing on rs-fMRI data. Then, a convolutional Autoencoder (CNN-AE) model with the proposed number of layers is used to extract features from rs-fMRI data. In the classification step, a new fuzzy method called interval type-2 fuzzy regression (IT2FR) is introduced and then optimized by genetic algorithm (GA), particle swarm optimization (PSO), and gray wolf optimization (GWO) techniques. Also, the results of IT2FR methods are compared with multilayer perceptron (MLP), k-nearest neighbors (KNN), support vector machine (SVM), random forest (RF), decision tree (DT), and adaptive neuro-fuzzy inference system (ANFIS) methods. The experiment results show that the IT2FR method with the GWO optimization algorithm has achieved satisfactory results compared to other classifier methods. Finally, the proposed classification technique was able to provide 72.71% accuracy.
... Traditionally, Random Forest (RF) classifier is utilized for multimodal classification, 15 and also discriminative sparse learning method deploys multimodal features to classify stages of AD. 16 These methods are computationally intensive and rely primarily on handcrafted features that are unappealing and difficult to obtain. Machine learning schemes [17][18][19][20] are found to be compatible to deal with these challenges. ...
... The procedure of the proposed logic is as follows: Here, when the iteration it is greater than or equal to 2 (it ≥ 2), the update takes place using two algorithms namely, DA and GWO. Specifically, when it =2 or when (19), and it is the average values of the position vectors of gray wolves α, β and γ. Otherwise, the random solution will be updated. ...
Article
Full-text available
Alzheimer's disease (AD) is a widespread neurodegenerative disease that causes 60–80% of all dementias and has a large economic impact in developed countries. For early‐stage AD detection, volumetric measures of magnetic resonance imaging (MRI) have proven to be a benchmark method. To detect potential cases, existing methods combine health records, neuropsychological testing, and MRI, although learning implementation is inconsistently used and has low sensitivity and specificity. Furthermore, numerous classification approaches for diagnosing AD have been suggested with differing complexity. Thus, we have introduced our novel AD diagnosis model with two main phases such as proposed feature extraction and classification. In the first phase, the gray‐level co‐occurrence matrix (GLCM), Haralick features as well as proposed geometric Haralick features known as geometric correlation and variance are extracted. In the second phase, an optimized deep convolutional neural network (DCNN) is utilized for classification. To make the prediction more accurate, the weight and the activation function of DCNN are optimally chosen by a new hybrid model termed as Combined Gray Wolf and Dragon Updating (CG‐DU). At last, the superiority of the adopted scheme is validated in terms of performance analysis, convergence analysis, box plot analysis, and computation time analysis. Especially, the proposed model achieves a mean accuracy of 0.98795, sensitivity of 0.98671, and specificity of 0.99429. Moreover, the computation time of the CG‐DU model is 2.92%, and 0.14% superior to existing GWO and DA methods respectively.
... DIAN constitutes a strong impact in AD research because it is remarkably helpful in the understanding of the disease's natural history (Morris et al., 2012). In the DIAN study, the classification rate between the mutation-carriers group and normal controls is approximately 80% using biomarkers with machine learning (Castillo-Barnes et al., 2020). The DIAN study is not only important for discovering disease trajectories (Luckett et al., 2021), but also for drug trials (Bateman et al., 2017). ...
... 76 PSEN1, 6 PSEN2, and 6 APP symptomatic mutation carriers were recruited to characterize neuroimaging biomarkers change in DIAN . Currently, 265 mutation carriers were included in DIAN, including 202 PSEN1, 22 PSEN2, and 43 APP mutation carriers (Castillo-Barnes et al., 2020;Luckett et al., 2021). In another France AD whole-exome sequencing study, three PSEN1 and one PSEN2 likely causative variants were identified (Nicolas et al., 2016). ...
Article
Full-text available
The strategies of classifying APP, PSEN1, and PSEN2 variants varied substantially in the previous studies. We aimed to re-evaluate these variants systematically according to the American college of medical genetics and genomics and the association for molecular pathology (ACMG-AMP) guidelines. In our study, APP, PSEN1, and PSEN2 variants were collected by searching Alzforum and PubMed database with keywords “PSEN1,” “PSEN2,” and “APP.” These variants were re-evaluated based on the ACMG-AMP guidelines. We compared the number of pathogenic/likely pathogenic variants of APP, PSEN1, and PSEN2. In total, 66 APP variants, 323 PSEN1 variants, and 63 PSEN2 variants were re-evaluated in our study. 94.91% of previously reported pathogenic variants were re-classified as pathogenic/likely pathogenic variants, while 5.09% of them were variants of uncertain significance (VUS). PSEN1 carried the most prevalent pathogenic/likely pathogenic variants, followed by APP and PSEN2. Significant statistically difference was identified among these three genes when comparing the number of pathogenic/likely pathogenic variants (P < 2.2 × 10–16). Most of the previously reported pathogenic variants were re-classified as pathogenic/likely pathogenic variants while the others were re-evaluated as VUS, highlighting the importance of interpreting APP, PSEN1, and PSEN2 variants with caution according to ACMG-AMP guidelines.
... In recent years, studies have shown that functional connectivity analysis based on fMRI modalities plays a significant role in the diagnosis of brain disorders such as SZ (Shoeibi et al. 2021a), Alzheimer's disease (AD) (Castillo-Barnes et al. 2020), epilepsy (Luckett et al. 2022;Anter et al. 2022), ADHD (Konrad and Eickhoff 2010), and BD (Highland and Zhou 2022). Brain disorders alter functional connectivity in the brain, which fMRI modalities can see. ...
Article
Full-text available
Nowadays, many people worldwide suffer from brain disorders, and their health is in danger. So far, numerous methods have been proposed for the diagnosis of Schizophrenia (SZ) and attention deficit hyperactivity disorder (ADHD), among which functional magnetic resonance imaging (fMRI) modalities are known as a popular method among physicians. This paper presents an SZ and ADHD intelligent detection method of resting-state fMRI (rs-fMRI) modality using a new deep learning method. The University of California Los Angeles dataset, which contains the rs-fMRI modalities of SZ and ADHD patients, has been used for experiments. The FMRIB software library toolbox first performed preprocessing on rs-fMRI data. Then, a convolutional Autoencoder model with the proposed number of layers is used to extract features from rs-fMRI data. In the classification step, a new fuzzy method called interval type-2 fuzzy regression (IT2FR) is introduced and then optimized by genetic algorithm, particle swarm optimization, and gray wolf optimization (GWO) techniques. Also, the results of IT2FR methods are compared with multilayer perceptron, k-nearest neighbors, support vector machine, random forest, and decision tree, and adaptive neuro-fuzzy inference system methods. The experiment results show that the IT2FR method with the GWO optimization algorithm has achieved satisfactory results compared to other classifier methods. Finally, the proposed classification technique was able to provide 72.71% accuracy.
... Indeed, while in a resubstitution scenario the upper bound is constant, during CV this need not be the case and we could find ourselves in scenarios of low effective learning capacity. All this makes RUB an optimal option due to the correction applied, especially for small sample sizes (common in neuroimaging), as it has been already tested [33][34][35]. ...
Article
Deep Learning (DL) predictions are uncertain; but how uncertain? Statistical inference estimates the probabilities of uncertainty from a sample drawn from a population. Assessing the statistical significance of accuracies reported by DL remains largely unexplored. A framework to do so would usefully support a range of applications, and in particular group classifications from neuroimages where, for operational reasons, sample sizes are necessary limited and thus often do not generalise well. We applied a random-effects inference based on a label permutation test to calculate the statistical significance of K-fold cross-validation (CV) from statistical power and Type-I error rates. Our hypothesis is that in low sample size scenarios, the use of resubstitution with upper bound correction (RUB) as a validation would mitigate the debate on the generalisation ability of DL models. The derived framework enables testing such generalisation ability of DL models as feature extraction methods. A combination of autoencoders and support vector machines as feature extraction and classification models is evaluated in a case-control analysis of Alzheimer’s disease with well-established outcomes. We found that RUB slightly outperforms K-fold CV as a validation method, especially estimating statistical power in the most heterogeneous samples. Therefore, we suggest RUB as potent and valid method for DL with neuroimages in terms of bias, variance and computational demand.
... Different types of diets have been studied to determine the nutritious intake desired for several diseases including Alzheimer's, but there is a need to automate this process to customise the dietary plans for each patient. Mediterranean, Dash, and MIND are some of the diets that have been studied in correlation with AD prevention and treatment [20,21]. Nutrition and dietary patterns can determine the probability of disease prevalence. ...
Article
Full-text available
Alzheimer’s disease (AD) has become a public health concern due to its misinterpretation with vascular dementia (VD) and mixed dementia Alzheimer’s disease (MXD). Therefore, an accurate differentiation of these diseases is essential for improving the treatment procedure. It has been seen that nutrition along with several other factors plays a role in the disease progression. Scientists are trying to find a solution using some machine learning (ML) techniques. The ML algorithms used for this purpose are neural networks, support vector machines, regression and many more. The current research is focused on understanding the extent of the application of machine learning tools in enhancing food management for patients with Alzheimer’s since there is no cure known for the same. A total of 100 patient data have been collected where the patients had AD, VD, and MXD. Their demographic data, dietary intake, Fazekas scores, and Hachinski scores were collected (independent variables) and analysed in IBM SPSS by considering the risk of development of AD, VD, and MXD as dependent variables. The findings showed that age is highly related ( p < 0.001 ) to the development of these three diseases and other demographics are not prioritized. Discussion of other available journal articles showed that nutritional intake, Fazekas scores, Hachinski scores, and gender are also indicators for predicting these diseases ( p < 0.001 ). Thus, this study concluded that age, gender, diet consumption, and Fazekas and Hachinski scores are important indicators for differentiating AD from other diseases, and ML can be used to create a custom nutrition plan based on the patient’s diet and stage of disease progression. Lastly, future scopes of ML have been explained in this paper.
... The linear SVM was evaluated with a nested k-fold CV approach. Nested CV is widely employed in the machine-learning analysis of neuro-imaging (12,(26)(27)(28)(29). Compared to simple CV, nested CV can reduce overfitting and limit optimistic biases, especially in relatively small samples (30,31). ...
Article
Full-text available
The detection of mutations in telomerase reverse transcriptase promoter (pTERT) is important since preoperative diagnosis of pTERT status helps with evaluating prognosis and determining the surgical strategy. Here, we aimed to establish a radiomics-based machine-learning algorithm and evaluated its performance with regard to the prediction of mutations in pTERT in patients with World Health Organization (WHO) grade II gliomas. In total, 164 patients with WHO grade II gliomas were enrolled in this retrospective study. We extracted a total of 1,293 radiomics features from multi-parametric magnetic resonance imaging scans. Elastic net (used for feature selection) and support vector machine with linear kernel were applied in nested 10-fold cross-validation loops. The predictive model was evaluated by receiver operating characteristic and precision-recall analyses. We performed an unpaired t-test to compare the posterior predictive probabilities among patients with differing pTERT statuses. We selected 12 valuable radiomics features using nested 10-fold cross-validation loops. The area under the curve (AUC) was 0.8446 (95% confidence interval [CI], 0.7735–0.9065) with an optimal summed value of sensitivity of 0.9355 (95% CI, 0.8802–0.9788) and specificity of 0.6197 (95% CI, 0.5071–0.7371). The overall accuracy was 0.7988 (95% CI, 0.7378–0.8598). The F1-score was 0.8406 (95% CI, 0.7684–0.902) with an optimal precision of 0.7632 (95% CI, 0.6818–0.8364) and recall of 0.9355 (95% CI, 0.8802–0.9788). Posterior probabilities of pTERT mutations were significantly different between patients with wild-type and mutant TERT promoters. Our findings suggest that a radiomics analysis with a machine-learning algorithm can be useful for predicting pTERT status in patients with WHO grade II glioma and may aid in glioma management.
... Traditional machine learning achieved excellent results on disease detections [25,26]. Convolutional neural network (CNN) is a new artificial neural network. ...
Article
Full-text available
(Aim) COVID-19 is an infectious disease spreading to the world this year. In this study, we plan to develop an artificial intelligence based tool to diagnose on chest CT images. (Method) On one hand, we extract features from a self-created convolutional neural network (CNN) to learn individual image-level representations. The proposed CNN employed several new techniques such as rank-based average pooling and multiple-way data augmentation. On the other hand, relation-aware representations were learnt from graph convolutional network (GCN). Deep feature fusion (DFF) was developed in this work to fuse individual image-level features and relation-aware features from both GCN and CNN, respectively. The best model was named as FGCNet. (Results) The experiment first chose the best model from eight proposed network models, and then compared it with 15 state-of-the-art approaches. (Conclusion) The proposed FGCNet model is effective and gives better performance than all 15 state-of-the-art methods. Thus, our proposed FGCNet model can assist radiologists to rapidly detect COVID-19 from chest CT images.
Article
Spatial normalization helps us to compare quantitatively two or more input brain scans. Although using an affine normalization approach preserves the anatomical structures, the neuroimaging field is more common to find works that make use of nonlinear transformations. The main reason is that they facilitate a voxel-wise comparison, not only when studying functional images but also when comparing MRI scans given that they fit better to a reference template. However, the amount of bias introduced by the nonlinear transformations can potentially alter the final outcome of a diagnosis especially when studying functional scans for neurological disorders like Parkinson's Disease. In this context, we have tried to quantify the bias introduced by the affine and the nonlinear spatial registration of FP-CIT SPECT volumes of healthy control subjects and patients with PD. For that purpose, we calculated the deformation fields of each participant and applied these deformation fields to a 3D-grid. As the space between the edges of small cubes comprising the grid change, we can quantify which parts from the brain have been enlarged, compressed or just remain the same. When the nonlinear approach is applied, scans from PD patients show a region near their striatum very similar in shape to that of healthy subjects. This artificially increases the interclass separation between patients with PD and healthy subjects as the local intensity is decreased in the latter region, and leads machine learning systems to biased results due to the artificial information introduced by these deformations.
Article
Alzheimer’s disease (AD) will become a global burden in the coming decades according to the latest statistical survey. How to effectively detect AD or MCI (mild cognitive impairment) using reliable biomarkers and robust machine learning methods has become a challenging problem. In this study, we propose a novel AD multiclass classification framework with embedding feature selection and fusion based on multimodal neuroimaging. The framework has three novel aspects: (1) An l_2,1-norm regularization term combined with the multiclass hinge loss is used to naturally select features across all the classes in each modality. (2) To fuse the complementary information contained in each modality, an l_p-norm regularization term is introduced to combine different kernels to perform multiple kernel learning to avoid a sparse kernel coefficient distribution, thereby effectively exploiting complementary modalities. (3) A theorem that transforms the multiclass hinge loss minimization problem using the l_2,1-norm and l_p-norm regularizations to a previous solvable optimization problem and its proof are given. Additionally, it is theoretically proved that the optimization process converges to the global optimum. Extensive comparison experiments and analysis support the promising performance of the proposed method.
Article
Several studies have recently suggested that amyloid Positron Emission Tomography (PET) data acquired immediately after the radiotracer injection provide information related to the brain metabolism, similar to that contained in 18F-Fluorodeoxyglucose (FDG) PET neuroimages. If corroborated, it would allow us to acquire information about brain injury and potential brain amyloid deposits in a single examination, using a dual-point protocol. In this work we assess the equivalence between early 18F-Florbetaben (FBB) PET and 18F-FDG PET data using multivariate approaches based on machine learning. In addition, we propose several systems based on data fusion that take advantage of the additional information provided by dual-point amyloid PET examinations. The proposed systems perform an initial dimensionality reduction of the data using a partial-least-square-based algorithm and then combine early and standard PET acquisitions using two approaches: multiple kernel learning (intermediate fusion) or an ensemble of two Support Vector Machine classifiers (late fusion). The proposed approaches were evaluated and compared with other fusion techniques using data from 43 subjects with cognitive impairments. They achieved a good trade-off between sensitivity and specificity and higher accuracy rates than systems based on single-modality approaches such as standard 18F-FBB PET data or 18F-FDG PET neuroimages.
Article
Full-text available
In 2011, the National Institute on Aging and Alzheimer's Association created separate diagnostic recommendations for the preclinical, mild cognitive impairment, and dementia stages of Alzheimer's disease. Scientific progress in the interim led to an initiative by the National Institute on Aging and Alzheimer's Association to update and unify the 2011 guidelines. This unifying update is labeled a “research framework” because its intended use is for observational and interventional research, not routine clinical care. In the National Institute on Aging and Alzheimer's Association Research Framework, Alzheimer's disease (AD) is defined by its underlying pathologic processes that can be documented by postmortem examination or in vivo by biomarkers. The diagnosis is not based on the clinical consequences of the disease (i.e., symptoms/signs) in this research framework, which shifts the definition of AD in living people from a syndromal to a biological construct. The research framework focuses on the diagnosis of AD with biomarkers in living persons. Biomarkers are grouped into those of β amyloid deposition, pathologic tau, and neurodegeneration [AT(N)]. This ATN classification system groups different biomarkers (imaging and biofluids) by the pathologic process each measures. The AT(N) system is flexible in that new biomarkers can be added to the three existing AT(N) groups, and new biomarker groups beyond AT(N) can be added when they become available. We focus on AD as a continuum, and cognitive staging may be accomplished using continuous measures. However, we also outline two different categorical cognitive schemes for staging the severity of cognitive impairment: a scheme using three traditional syndromal categories and a six-stage numeric scheme. It is important to stress that this framework seeks to create a common language with which investigators can generate and test hypotheses about the interactions among different pathologic processes (denoted by biomarkers) and cognitive symptoms. We appreciate the concern that this biomarker-based research framework has the potential to be misused. Therefore, we emphasize, first, it is premature and inappropriate to use this research framework in general medical practice. Second, this research framework should not be used to restrict alternative approaches to hypothesis testing that do not use biomarkers. There will be situations where biomarkers are not available or requiring them would be counterproductive to the specific research goals (discussed in more detail later in the document). Thus, biomarker-based research should not be considered a template for all research into age-related cognitive impairment and dementia; rather, it should be applied when it is fit for the purpose of the specific research goals of a study. Importantly, this framework should be examined in diverse populations. Although it is possible that β-amyloid plaques and neurofibrillary tau deposits are not causal in AD pathogenesis, it is these abnormal protein deposits that define AD as a unique neurodegenerative disease among different disorders that can lead to dementia. We envision that defining AD as a biological construct will enable a more accurate characterization and understanding of the sequence of events that lead to cognitive impairment that is associated with AD, as well as the multifactorial etiology of dementia. This approach also will enable a more precise approach to interventional trials where specific pathways can be targeted in the disease process and in the appropriate people.
Article
Full-text available
Dominantly-inherited Alzheimer's disease is widely hoped to hold the key to developing interventions for sporadic late onset Alzheimer's disease. We use emerging techniques in generative data-driven disease progression modelling to characterize dominantly-inherited Alzheimer's disease progression with unprecedented resolution, and without relying upon familial estimates of years until symptom onset. We retrospectively analysed biomarker data from the sixth data freeze of the Dominantly Inherited Alzheimer Network observational study, including measures of amyloid proteins and neurofibrillary tangles in the brain, regional brain volumes and cortical thicknesses, brain glucose hypometabolism, and cognitive performance from the Mini-Mental State Examination (all adjusted for age, years of education, sex, and head size, as appropriate). Data included 338 participants with known mutation status (211 mutation carriers in three subtypes: 163 PSEN1, 17 PSEN2, and 31 APP) and a baseline visit (age 19-66; up to four visits each, 1.1 ± 1.9 years in duration; spanning 30 years before, to 21 years after, parental age of symptom onset). We used an event-based model to estimate sequences of biomarker changes from baseline data across disease subtypes (mutation groups), and a differential equation model to estimate biomarker trajectories from longitudinal data (up to 66 mutation carriers, all subtypes combined). The two models concur that biomarker abnormality proceeds as follows: amyloid deposition in cortical then subcortical regions (∼24 ± 11 years before onset); phosphorylated tau (17 ± 8 years), tau and amyloid-β changes in cerebrospinal fluid; neurodegeneration first in the putamen and nucleus accumbens (up to 6 ± 2 years); then cognitive decline (7 ± 6 years), cerebral hypometabolism (4 ± 4 years), and further regional neurodegeneration. Our models predicted symptom onset more accurately than predictions that used familial estimates: root mean squared error of 1.35 years versus 5.54 years. The models reveal hidden detail on dominantly-inherited Alzheimer's disease progression, as well as providing data-driven systems for fine-grained patient staging and prediction of symptom onset with great potential utility in clinical trials.
Article
Full-text available
Worldwide there are over 46 million people living with dementia, and this number is expected to double every 20 years reaching about 131 million by 2050. The cost to the community and government health systems, as well as the stress on families and carers is incalculable. Over three decades of research into this disease have been undertaken by several research groups in Australia, including work by our original research group in Western Australia which was involved in the discovery and sequencing of the amyloid-β peptide (also known as Aβ or A4 peptide) extracted from cerebral amyloid plaques. This review discusses the journey from the discovery of the Aβ peptide in Alzheimer's disease (AD) brain to the establishment of pre-clinical AD using PET amyloid tracers, a method now serving as the gold standard for developing peripheral diagnostic approaches in the blood and the eye. The latter developments for early diagnosis have been largely achieved through the establishment of the Australian Imaging Biomarker and Lifestyle research group that has followed 1,100 Australians for 11 years. AIBL has also been instrumental in providing insight into the role of the major genetic risk factor apolipoprotein E ɛ4, as well as better understanding the role of lifestyle factors particularly diet, physical activity and sleep to cognitive decline and the accumulation of cerebral Aβ.
Article
Full-text available
Protein misfolding and aggregation is a central feature of several neurodegenerative disorders including Alzheimer's disease (AD), in which assemblies of amyloid β (Aβ) peptides accumulate in the brain in the form of parenchymal and/or vascular amyloid. A widely accepted concept is that AD is characterized by distinct clinical and neuropathological phenotypes. Recent studies revealed that Aβ assemblies might have structural differences among AD brains and that such pleomorphic assemblies can correlate with distinct disease phenotypes. We found that in both sporadic and inherited forms of AD, amyloid aggregates differ in the biochemical composition of Aβ species. These differences affect the physicochemical properties of Aβ assemblies including aggregation kinetics, resistance to degradation by proteases and seeding ability. Aβ-amyloidosis can be induced and propagated in animal models by inoculation of brain extracts containing aggregated Aβ. We found that brain homogenates from AD patients with different molecular profiles of Aβ are able to induce distinct patterns of Aβ-amyloidosis when injected into mice. Overall these data suggest that the assembly of mixtures of Aβ peptides into different Aβ seeds leads to the formation of distinct subtypes of amyloid having distinctive physicochemical and biological properties which result in the generation of distinct AD molecular subgroups.
Article
Full-text available
Autosomal dominant Alzheimer's disease (ADAD) is a small subset of Alzheimer's disease that is genetically determined with 100% penetrance. It provides a valuable window into studying the course of pathologic processes that leads to dementia. Arterial spin labeling (ASL) MRI is a potential AD imaging marker that non-invasively measures cerebral perfusion. In this study, we investigated the relationship of cerebral blood flow measured by pseudo-continuous ASL (pCASL) MRI with measures of cerebral metabolism (FDG PET) and amyloid deposition (Pittsburgh Compound B (PiB) PET). Thirty-one participants at risk for ADAD (age 39 ± 13 years, 19 females) were recruited into this study, and 21 of them received both MRI and FDG and PiB PET scans. Considerable variability was observed in regional correlations between ASL-CBF and FDG across subjects. Both regional hypo-perfusion and hypo-metabolism were associated with amyloid deposition. Cross-sectional analyses of each biomarker as a function of the estimated years to expected dementia diagnosis indicated an inverse relationship of both perfusion and glucose metabolism with amyloid deposition during AD development. These findings indicate that neurovascular dysfunction is associated with amyloid pathology, and also indicate that ASL CBF may serve as a sensitive early biomarker for AD. The direct comparison among the three biomarkers provides complementary information for understanding the pathophysiological process of AD.
Article
Full-text available
Background: The causes of phenotypic heterogeneity in familial Alzheimer's disease with autosomal dominant inheritance are not well understood. We aimed to characterise clinical phenotypes and genetic associations with APP and PSEN1 mutations in symptomatic autosomal dominant familial Alzheimer's disease (ADAD). Methods: We retrospectively analysed genotypic and phenotypic data (age at symptom onset, initial cognitive or behavioural symptoms, and presence of myoclonus, seizures, pyramidal signs, extrapyramidal signs, and cerebellar signs) from all individuals with ADAD due to APP or PSEN1 mutations seen at the Dementia Research Centre in London, UK. We examined the frequency of presenting symptoms and additional neurological features, investigated associations with age at symptom onset, APOE genotype, and mutation position, and explored phenotypic differences between APP and PSEN1 mutation carriers. The proportion of individuals presenting with various symptoms was analysed with descriptive statistics, stratified by mutation type. Findings: Between July 1, 1987, and Oct 31, 2015, age at onset was recorded for 213 patients (168 with PSEN1 mutations and 45 with APP mutations), with detailed history and neurological examination findings available for 121 (85 with PSEN1 mutations and 36 with APP mutations). We identified 38 different PSEN1 mutations (four novel) and six APP mutations (one novel). Age at onset differed by mutation, with a younger onset for individuals with PSEN1 mutations than for those with APP mutations (mean age 43·6 years [SD 7·2] vs 50·4 years [SD 5·2], respectively, p<0·0001); within the PSEN1 group, 72% of age at onset variance was explained by the specific mutation. A cluster of five mutations with particularly early onset (mean age at onset <40 years) involving PSEN1's first hydrophilic loop suggests critical functional importance of this region. 71 (84%) individuals with PSEN1 mutations and 35 (97%) with APP mutations presented with amnestic symptoms, making atypical cognitive presentations significantly more common in PSEN1 mutation carriers (n=14; p=0·037). Myoclonus and seizures were the most common additional neurological features; individuals with myoclonus (40 [47%] with PSEN1 mutations and 12 [33%] with APP mutations) were significantly more likely to develop seizures (p=0·001 for PSEN1; p=0·036 for APP), which affected around a quarter of the patients in each group (20 [24%] and nine [25%], respectively). A number of patients with PSEN1 mutations had pyramidal (21 [25%]), extrapyramidal (12 [14%]), or cerebellar (three [4%]) signs. Interpretation: ADAD phenotypes are heterogeneous, with both age at onset and clinical features being influenced by mutation position as well as causative gene. This highlights the importance of considering genetic testing in young patients with dementia and additional neurological features in order to appropriately diagnose and treat their symptoms, and of examining different mutation types separately in future research. Funding: Medical Research Council and National Institute for Health Research.
Article
Full-text available
Introduction: The Dominantly Inherited Alzheimer Network Trial Unit (DIAN-TU) is an adaptive platform trial testing multiple drugs to slow or prevent the progression of Alzheimer's disease in autosomal dominant Alzheimer's disease (ADAD) families. With completion of enrollment of the first two drug arms, the DIAN-TU now plans to add new drugs to the platform, designated as the Next Generation (NexGen) prevention trial. Methods: In collaboration with ADAD families, philanthropic organizations, academic leaders, the DIAN-TU Pharma Consortium, the National Institutes of Health, and regulatory colleagues, the DIAN-TU developed innovative clinical study designs for the DIAN-TU NexGen prevention trial. Results: Our expanded trial toolbox consists of a disease progression model for ADAD, primary end point DIAN-TU cognitive performance composite, biomarker development, self-administered cognitive assessments, adaptive dose adjustments, and blinded data collection through the last participant completion. Conclusion: These steps represent elements to improve efficacy of the adaptive platform trial and a continued effort to optimize prevention and treatment trials in ADAD.
Article
Full-text available
Amyloid imaging plays an important role in the research and diagnosis of dementing disorders. Substantial variation in quantitative methods to measure brain amyloid burden exists in the field. The aim of this work is to investigate the impact of methodological variations to the quantification of amyloid burden using data from the Dominantly Inherited Alzheimer's Network (DIAN), an autosomal dominant Alzheimer's disease population. Cross-sectional and longitudinal [11C]-Pittsburgh Compound B (PiB) PET imaging data from the DIAN study were analyzed. Four candidate reference regions were investigated for estimation of brain amyloid burden. A regional spread function based technique was also investigated for the correction of partial volume effects. Cerebellar cortex, brain-stem, and white matter regions all had stable tracer retention during the course of disease. Partial volume correction consistently improves sensitivity to group differences and longitudinal changes over time. White matter referencing improved statistical power in the detecting longitudinal changes in relative tracer retention; however, the reason for this improvement is unclear and requires further investigation. Full dynamic acquisition and kinetic modeling improved statistical power although it may add cost and time. Several technical variations to amyloid burden quantification were examined in this study. Partial volume correction emerged as the strategy that most consistently improved statistical power for the detection of both longitudinal changes and across-group differences. For the autosomal dominant Alzheimer's disease population with PiB imaging, utilizing brainstem as a reference region with partial volume correction may be optimal for current interventional trials. Further investigation of technical issues in quantitative amyloid imaging in different study populations using different amyloid imaging tracers is warranted.
Article
Full-text available
Alzheimer disease (AD) represents a genetically heterogeneous entity. To elucidate neuropathologic features of autosomal dominant AD ([ADAD] due to PSEN1, APP, or PSEN2 mutations), we compared hallmark AD pathologic findings in 60 cases of ADAD and 120 cases of sporadic AD matched for sex, race, ethnicity, and disease duration. Greater degrees of neuritic plaque and neurofibrillary tangle formation and cerebral amyloid angiopathy (CAA) were found in ADAD (p values < 0.01). Moderate to severe CAA was more prevalent in ADAD (63.3% vs. 39.2%, p = 0.003), and persons with PSEN1 mutations beyond codon 200 had higher average Braak scores and severity and prevalence of CAA than those with mutations before codon 200. Lewy body pathology was less extensive in ADAD but was present in 27.1% of cases. We also describe a novel pathogenic PSEN1 mutation (P267A). The finding of more severe neurofibrillary pathology and CAA in ADAD, particularly in carriers of PSEN1 mutations beyond codon 200, warrants consideration when designing trials to treat or prevent ADAD. The finding of Lewy body pathology in a substantial minority of ADAD cases supports the assertion that development of Lewy bodies may be in part driven by abnormal β-amyloid protein precursor processing.
Article
Full-text available
Objective: We examined the diagnostic value of subjective memory complaints (SMCs) assessed with a single item in a large cross-sectional cohort consisting of families with autosomal dominant Alzheimer's disease (ADAD) participating in the Dominantly Inherited Alzheimer Network (DIAN). Methods: The baseline sample of 183 mutation carriers (MCs) and 117 noncarriers (NCs) was divided according to Clinical Dementia Rating (CDR) scale into preclinical (CDR 0; MCs: n = 107; NCs: n = 109), early symptomatic (CDR 0.5; MCs: n = 48; NCs: n = 8), and dementia stage (CDR ≥ 1; MCs: n = 28; NCs: n = 0). These groups were subdivided by the presence or absence of SMCs. Results: At CDR 0, SMCs were present in 12.1% of MCs and 9.2% of NCs (P = 0.6). At CDR 0.5, SMCs were present in 66.7% of MCs and 62.5% of NCs (P = 1.0). At CDR ≥ 1, SMCs were present in 96.4% of MCs. SMCs in MCs were significantly associated with CDR, logical memory scores, Geriatric Depression Scale, education, and estimated years to onset. Conclusions: The present study shows that SMCs assessed by a single-item scale have no diagnostic value to identify preclinical ADAD in asymptomatic individuals. These results demonstrate the need of further improvement of SMC measures that should be examined in large clinical trials.
Article
Full-text available
Three sets of research criteria are available for diagnosis of Alzheimer's disease in subjects with mild cognitive impairment: the International Working Group-1, International Working Group-2, and National Institute of Aging-Alzheimer Association criteria. We compared the prevalence and prognosis of Alzheimer's disease at the mild cognitive impairment stage according to these criteria. Subjects with mild cognitive impairment (n = 1607), 766 of whom had both amyloid and neuronal injury markers, were recruited from 13 cohorts. We used cognitive test performance and available biomarkers to classify subjects as prodromal Alzheimer's disease according to International Working Group-1 and International Working Group-2 criteria and in the high Alzheimer's disease likelihood group, conflicting biomarker groups (isolated amyloid pathology or suspected non-Alzheimer pathophysiology), and low Alzheimer's disease likelihood group according to the National Institute of Ageing-Alzheimer Association criteria. Outcome measures were the proportion of subjects with Alzheimer's disease at the mild cognitive impairment stage and progression to Alzheimer's disease-type dementia. We performed survival analyses using Cox proportional hazards models. According to the International Working Group-1 criteria, 850 (53%) subjects had prodromal Alzheimer's disease. Their 3-year progression rate to Alzheimer's disease-type dementia was 50% compared to 21% for subjects without prodromal Alzheimer's disease. According to the International Working Group-2 criteria, 308 (40%) subjects had prodromal Alzheimer's disease. Their 3-year progression rate to Alzheimer's disease-type dementia was 61% compared to 22% for subjects without prodromal Alzheimer's disease. According to the National Institute of Ageing-Alzheimer Association criteria, 353 (46%) subjects were in the high Alzheimer's disease likelihood group, 49 (6%) in the isolated amyloid pathology group, 220 (29%) in the suspected non-Alzheimer pathophysiology group, and 144 (19%) in the low Alzheimer's disease likelihood group. The 3-year progression rate to Alzheimer's disease-type dementia was 59% in the high Alzheimer's disease likelihood group, 22% in the isolated amyloid pathology group, 24% in the suspected non-Alzheimer pathophysiology group, and 5% in the low Alzheimer's disease likelihood group. Our findings support the use of the proposed research criteria to identify Alzheimer's disease at the mild cognitive impairment stage. In clinical settings, the use of both amyloid and neuronal injury markers as proposed by the National Institute of Ageing-Alzheimer Association criteria offers the most accurate prognosis. For clinical trials, selection of subjects in the National Institute of Ageing-Alzheimer Association high Alzheimer's disease likelihood group or the International Working Group-2 prodromal Alzheimer's disease group could be considered. © The Author (2015). Published by Oxford University Press on behalf of the Guarantors of Brain. All rights reserved. For Permissions, please email: journals.permissions@oup.com.
Article
Full-text available
Since the original publication describing the illness in 1907, the genetic understanding of Alzheimer's disease (AD) has advanced such that it is now clear that it is a genetically heterogeneous condition, the subtypes of which may not uniformly respond to a given intervention. It is therefore critical to characterize the clinical and preclinical stages of AD subtypes, including the rare autosomal dominant forms caused by known mutations in the PSEN1, APP, and PSEN2 genes that are being studied in the Dominantly Inherited Alzheimer Network study and its associated secondary prevention trial. Similar efforts are occurring in an extended Colombian family with a PSEN1 mutation, in APOE ε4 homozygotes, and in Down syndrome. Despite commonalities in the mechanisms producing the AD phenotype, there are also differences that reflect specific genetic origins. Treatment modalities should be chosen and trials designed with these differences in mind. Ideally, the varying pathological cascades involved in the different subtypes of AD should be defined so that both areas of overlap and of distinct differences can be taken into account. At the very least, clinical trials should determine the influence of known genetic factors in post hoc analyses.
Article
Full-text available
Importance Autosomal dominant Alzheimer disease (ADAD) is caused by rare genetic mutations in 3 specific genes in contrast to late-onset Alzheimer disease (LOAD), which has a more polygenetic risk profile.Objective To assess the similarities and differences in functional connectivity changes owing to ADAD and LOAD.Design, Setting, and Participants We analyzed functional connectivity in multiple brain resting state networks (RSNs) in a cross-sectional cohort of participants with ADAD (n = 79) and LOAD (n = 444), using resting-state functional connectivity magnetic resonance imaging at multiple international academic sites.Main Outcomes and Measures For both types of AD, we quantified and compared functional connectivity changes in RSNs as a function of dementia severity measured by the Clinical Dementia Rating Scale. In ADAD, we qualitatively investigated functional connectivity changes with respect to estimated years from onset of symptoms within 5 RSNs.Results A decrease in functional connectivity with increasing Clinical Dementia Rating scores were similar for both LOAD and ADAD in multiple RSNs. Ordinal logistic regression models constructed in one type of Alzheimer disease accurately predicted clinical dementia rating scores in the other, further demonstrating the similarity of functional connectivity loss in each disease type. Among participants with ADAD, functional connectivity in multiple RSNs appeared qualitatively lower in asymptomatic mutation carriers near their anticipated age of symptom onset compared with asymptomatic mutation noncarriers.Conclusions and Relevance Resting-state functional connectivity magnetic resonance imaging changes with progressing AD severity are similar between ADAD and LOAD. Resting-state functional connectivity magnetic resonance imaging may be a useful end point for LOAD and ADAD therapy trials. Moreover, the disease process of ADAD may be an effective model for the LOAD disease process.
Conference Paper
Full-text available
We review accuracy estimation methods and compare the two most common methods: crossvalidation and bootstrap. Recent experimental results on arti cial data and theoretical results in restricted settings have shown that for selecting a good classi er from a set of classiers (model selection), ten-fold cross-validation may be better than the more expensive leaveone-out cross-validation. We report on a largescale experiment|over half a million runs of C4.5 and a Naive-Bayes algorithm|to estimate the e ects of di erent parameters on these algorithms on real-world datasets. For crossvalidation, we vary the number of folds and whether the folds are strati ed or not � for bootstrap, we vary the number of bootstrap samples. Our results indicate that for real-word datasets similar to ours, the best method to use for model selection is ten-fold strati ed cross validation, even if computation power allows using more folds. 1
Article
Full-text available
Clinicopathological evidence suggests that the pathology of Alzheimer's disease (AD) begins many years before the appearance of cognitive symptoms. Biomarkers are required to identify affected individuals during this asymptomatic ("preclinical") stage to permit intervention with potential disease-modifying therapies designed to preserve normal brain function. Studies of families with autosomal-dominant AD (ADAD) mutations provide a unique and powerful means to investigate AD biomarker changes during the asymptomatic period. In this biomarker study, we collected cerebrospinal fluid (CSF), plasma, and in vivo amyloid imaging cross-sectional data at baseline in individuals from ADAD families enrolled in the Dominantly Inherited Alzheimer Network. Our study revealed reduced concentrations of CSF amyloid-β1-42 (Aβ1-42) associated with the presence of Aβ plaques, and elevated concentrations of CSF tau, ptau181 (phosphorylated tau181), and VILIP-1 (visinin-like protein-1), markers of neurofibrillary tangles and neuronal injury/death, in asymptomatic mutation carriers 10 to 20 years before their estimated age at symptom onset (EAO) and before the detection of cognitive deficits. When compared longitudinally, however, the concentrations of CSF biomarkers of neuronal injury/death within individuals decreased after their EAO, suggesting a slowing of acute neurodegenerative processes with symptomatic disease progression. These results emphasize the importance of longitudinal, within-person assessment when modeling biomarker trajectories across the course of the disease. If corroborated, this pattern may influence the definition of a positive neurodegenerative biomarker outcome in clinical trials.
Article
Full-text available
A candidate gene for the chromosome 1 Alzheimer's disease (AD) locus was identified (STM2). The predicted amino acid sequence for STM2 is homologous to that of the recently cloned chromosome 14 AD gene (S182). A point mutation in STM2, resulting in the substitution of an isoleucine for an asparagine (N141l), was identified in affected people from Volga German AD kindreds. This N141l mutation occurs at an amino acid residue that is conserved in human S182 and in the mouse S182 homolog. The presence of missense mutations in AD subjects in two highly similar genes strongly supports the hypothesis that mutations in both are pathogenic.
Article
Full-text available
Significance Beta-amyloid plaque accumulation, glucose hypometabolism, and neuronal atrophy are hallmarks of Alzheimer’s disease. However, the regional ordering of these biomarkers prior to dementia remains untested. In a cohort with Alzheimer’s disease mutations, we performed an integrated whole-brain analysis of three major imaging techniques: amyloid PET, [ ¹⁸ F]fluro-deoxyglucose PET, and structural MRI. We found that most gray-matter structures with amyloid plaques later have hypometabolism followed by atrophy. Critically, however, not all regions lose metabolic function, and not all regions atrophy, even when there is significant amyloid deposition. These regional disparities have important implications for clinical trials of disease-modifying therapies.
Article
Full-text available
We describe a comprehensive linear approach to the problem of imaging brain activity with high temporal as well as spatial resolution based on combining EEG and MEG data with anatomical constraints derived from MRI images. The "inverse problem" of estimating the distribution of dipole strengths over the cortical surface is highly underdetermined, even given closely spaced EEG and MEG recordings. We have obtained much better solutions to this problem by explicitly incorporating both local cortical orientation as well as spatial covariance of sources and sensors into our formulation. An explicit polygonal model of the cortical manifold is first constructed as follows: (1) slice data in three orthogonal planes of section (needle-shaped voxels) are combined with a linear deblurring technique to make a single high-resolution 3-D image (cubic voxels), (2) the image is recursively flood-filled to determine the topology of the gray-white matter border, and (3) the resulting continuous surface is refined by relaxing it against the original 3-D gray-scale image using a deformable template method, which is also used to computationally flatten the cortex for easier viewing. The explicit solution to an error minimization formulation of an optimal inverse linear operator (for a particular cortical manifold, sensor placement, noise and prior source covariance) gives rise to a compact expression that is practically computable for hundreds of sensors and thousands of sources. The inverse solution can then be weighted for a particular (averaged) event using the sensor covariance for that event. Model studies suggest that we may be able to localize multiple cortical sources with spatial resolution as good as PET with this technique, while retaining a much finer grained picture of activity over time.
Article
Full-text available
Mutations in the presenilin1 (PSEN1) and amyloid beta (A4) precursor protein (APP) genes account for the majority of cases of autosomal dominantly inherited Alzheimer's disease (AD). We wished to assess and compare the patterns of cerebral loss produced by these two groups of mutations. Volumetric magnetic resonance imaging and neuropsychological assessments were performed in individuals with clinical AD carrying mutations in the APP (n = 10) and PSEN1 (n = 18) genes and in healthy controls (n = 18). Voxel-based morphometry (VBM), cortical thickness, and region of interest analyses were performed. Mini-Mental State Examination scores were similar in the two disease groups suggesting similar levels of disease severity. There was evidence that APP subjects have smaller hippocampal volume compared with PSEN1 subjects (p = 0.007), and weak evidence that they have larger whole-brain and grey matter volumes (both p = 0.07). Although there was no evidence of statistically significant differences between APP and PSEN1 in VBM or cortical thickness analyses, effect maps were suggestive of APP subjects having more medial temporal lobe atrophy and conversely PSEN1 subjects showing more neocortical loss. Neuropsychological data were consistent with these regional differences and suggested greater memory deficits in the APP patients and greater impairment in non-memory domains in the PSEN1 group, although these differences were not statistically significant. We conclude that the mechanisms by which APP and PSEN1 mutations cause neuronal loss may differ which furthers our understanding of the neuropathology underlying AD and may inform future therapeutic strategies and trial designs.
Article
Full-text available
Autosomal-dominant Alzheimer disease (ADAD) is a genetic disorder caused by mutations in Amyloid Precursor Protein (APP) or Presenilin (PSEN) genes. Studies from families with ADAD have been critical to support the amyloid cascade hypothesis of Alzheimer disease (AD), the basis for the current development of amyloid-based disease-modifying therapies in sporadic AD (SAD). However, whether the pathological changes in APP processing in the CNS in ADAD are similar to those observed in SAD remains unclear. In this study, we measured β-site APP-cleaving enzyme (BACE) protein levels and activity, APP and APP C-terminal fragments in brain samples from subjects with ADAD carrying APP or PSEN1 mutations (n = 18), patients with SAD (n = 27) and age-matched controls (n = 22). We also measured sAPPβ and BACE protein levels, as well as BACE activity, in CSF from individuals carrying PSEN1 mutations (10 mutation carriers and 7 non-carrier controls), patients with SAD (n = 32) and age-matched controls (n = 11). We found that in the brain, the pattern in ADAD was characterized by an increase in APP β-C-terminal fragment (β-CTF) levels despite no changes in BACE protein levels or activity. In contrast, the pattern in SAD in the brain was mainly characterized by an increase in BACE levels and activity, with less APP β-CTF accumulation than ADAD. In the CSF, no differences were found between groups in BACE activity or expression or sAPPβ levels. Taken together, these data suggest that the physiopathological events underlying the chronic Aβ production/clearance imbalance in SAD and ADAD are different. These differences should be considered in the design of intervention trials in AD.
Article
In this paper we derive practical and novel upper bounds for the resubstitution error estimate by assessing the number of linear decision functions within the problem of pattern recognition in neuroimaging. Linear classifiers and regressors have been considered in many fields, where the number of predictors far exceeds the number of training samples available, to overcome the limitations of high complexity models in terms of computation, interpretability and overfitting. Typically in neuroimaging this is the rule rather than the exception, since the dimensionality of each observation (millions of voxels) in relation to the number of available samples (hundred of scans) implies a high risk of overfitting. Based on classical combinatorial geometry, we estimate the number of hyperplanes or linear decision rules and the corresponding distribution-independent performance bounds, comparing it to those obtained by the use of the VC-dimension concept. Experiments on synthetic and neuroimaging data demonstrate the performance of resubstitution error estimators, which are often overlooked in heterogeneous scenarios where their performance is similar to that obtained by cross-validation methods.
Article
In the recent 5 years (2014–2018), there has been growing interest in the use of machine learning (ML) techniques to explore image diagnosis and prognosis of therapeutic lesion changes within the area of neuroradiology. However, to date, the majority of research trend and current status have not been clearly illuminated in the neuroradiology field. More than 1000 papers have been published during the past 5 years on subject classification and prediction focused on multiple brain disorders. We provide a survey of 209 papers in this field with a focus on top ten active areas of research; i.e., Alzheimer’s disease/mild cognitive impairment, brain tumor; schizophrenia, depressive disorders, Parkinson’s disease, attention-deficit hyperactivity disorder, autism spectrum disease, epilepsy, multiple sclerosis, stroke, and traumatic brain injury. Detailed information of these studies, such as ML methods, sample size, type of inputted features and reported accuracy, are summarized. This paper reviews the evidences, current limitations and status of studies using ML to assess brain disorders in neuroimaging data. The main bottleneck of this research field is still the limited sample size, which could be potentially addressed by modern data sharing models, such as ADNI.
Article
Background Alzheimer's disease (AD) is the most common cause of dementia in the elderly and affects approximately 30 million individuals worldwide. Mild cognitive impairment (MCI) is very frequently a prodromal phase of AD, and existing studies have suggested that people with MCI tend to progress to AD at a rate of about 10–15% per year. However, the ability of clinicians and machine learning systems to predict AD based on MRI biomarkers at an early stage is still a challenging problem that can have a great impact in improving treatments. Method The proposed system, developed by the SiPBA-UGR team for this challenge, is based on feature standardization, ANOVA feature selection, partial least squares feature dimension reduction and an ensemble of One vs. Rest random forest classifiers. With the aim of improving its performance when discriminating healthy controls (HC) from MCI, a second binary classification level was introduced that reconsiders the HC and MCI predictions of the first level. Results The system was trained and evaluated on an ADNI datasets that consist of T1-weighted MRI morphological measurements from HC, stable MCI, converter MCI and AD subjects. The proposed system yields a 56.25% classification score on the test subset which consists of 160 real subjects. Comparison with existing method(s) The classifier yielded the best performance when compared to: (i) One vs. One (OvO), One vs. Rest (OvR) and error correcting output codes (ECOC) as strategies for reducing the multiclass classification task to multiple binary classification problems, (ii) support vector machines, gradient boosting classifier and random forest as base binary classifiers, and (iii) bagging ensemble learning. Conclusions A robust method has been proposed for the international challenge on MCI prediction based on MRI data. The system yielded the second best performance during the competition with an accuracy rate of 56.25% when evaluated on the real subjects of the test set.
Article
Emerging evidence supports a role for innate immunity and microglia in Alzheimer’s disease (AD) pathophysiology. However, no marker related to microglia has been included in the temporal evolution models of AD. TREM2 is a transmembrane protein involved in innate immunity and is selectively expressed by microglia and genetically linked to AD and other neurodegenerative disorders. Its ectodomain is released by proteolysis as a soluble variant (sTREM2) and can be detected in the cerebrospinal fluid (CSF). In patients with autosomal dominant AD, we tested how many years before the expected symptom onset did CSF sTREM2 increase in mutation carriers (MCs) compared to noncarriers (NCs). We also determined the temporal sequence of changes in CSF sTREM2 and markers for amyloid deposition and neurodegeneration as well as cognitive performance. We included 218 participants consisting of 127 MC and 91 NC siblings from the Dominantly Inherited Alzheimer Network. We observed that CSF sTREM2 increased in MCs compared to NCs 5 years before the expected symptom onset and this difference remained significant until 5 years after the expected symptom onset. Changes in CSF sTREM2 occurred after alterations were observed in markers for brain amyloidosis and neuronal injury. We propose that microglial activation occurs several years before the expected symptom onset, but after amyloidosis and neuronal injury have already occurred.
Article
Background: Autosomal dominant familial Alzheimer's disease (ADAD) is a rare disorder with non-amnestic neurological symptoms in some clinical presentations. We aimed to compile and compare data from symptomatic participants in the Dominantly Inherited Alzheimer Network observational study (DIAN-OBS) with those reported in the literature to estimate the prevalences of non-amnestic neurological symptoms in participants with ADAD. Methods: We prospectively collected data from the DIAN-OBS database, which recruited participants from study centres in the USA, Europe, and Australia, between Feb 29, 2008, and July 1, 2014. We also did a systematic review of publications to extract individual-level clinical data for symptomatic participants with ADAD. We used data for age of onset (from first report of cognitive decline), disease course from onset to death, and the presence of 13 neurological findings that have been reported in association with ADAD. Using multivariable linear regression, we investigated the prevalences of various non-amnestic neurological symptoms and the contributions of age of onset and specific mutation type on symptoms. Findings: The DIAN-OBS dataset included 107 individuals with detailed clinical data (forming the DIAN-OBS cohort). Our systematic review yielded 188 publications reporting on 1228 symptomatic individuals, with detailed neurological examination descriptions available for 753 individuals (forming the published data cohort). The most prevalent non-amnestic cognitive manifestations in participants in the DIAN-OBS cohort were those typical of mild to moderate Alzheimer's disease, including visual agnosia (55·1%, 95% CI 45·7-64·6), aphasia (57·9%, 48·6-67·3), and behavioural changes (61·7%, 51·5-70·0). Non-amnestic cognitive manifestations were less prevalent in the published data cohort (eg, visual agnosia [5·6%, 3·9-7·2], aphasia [23·0%, 20·0-26·0], and behavioural changes [31·7%, 28·4-35·1]). Prevalence of non-cognitive neurological manifestations in the DIAN-OBS cohort was low, including myoclonus and spasticity (9·3%, 95% CI 3·8-15·0), and seizures (2·8%, 0·5-5·9) and moderate for parkinsonism (11·2%, 5·3-17·1). By constrast, prevalence was higher in the published data cohort for myoclonus and spasticity (19·4%, 16·6-22·2 and 15·0%, 12·5-17·6, respectively), parkinsonism (12·5%, 10·1-15·0), and seizures (20·3%, 17·4-23·2). In an analysis of the published data cohort, ischaemic stroke was more prevalent at older ages of onset of symptoms of ADAD (odds ratio 1·09 per 1 year increase in age of onset, 95% CI 1·04-1·14, p=0·0003); and motor symptoms were more common at younger age of onset (myoclonus 0·93, 0·90-0·97, p=0·0007; seizures 0·95, 0·92-0·98, p=0·0018; corticobulbar deficits 0·91, 0·86-0·96, p=0·0012; and cerebellar ataxia 0·82, 0·74-0·91, p=0·0002). In the DIAN-OBS cohort, non-cognitive symptoms were more common at more severe stages of disease. Interpretation: The non-cognitive clinical manifestations of Alzheimer's disease seem to affect a small proportion of participants with mild to moderate ADAD, and are probably influenced by disease severity, environmental, and genetic factors. When evaluating patients with potential ADAD, clinicians should note that cognitive symptoms typical of sporadic Alzheimer's disease are the most consistent finding, with some patients manifesting non-cognitive neurological symptoms. Future work is needed to determine the environmental and genetic factors that cause these neurological symptoms. Funding: National Institutes of Health and German Center for Neurodegenerative Diseases.
Article
Objective: White matter hyperintensities(WMH) are areas of increased signal on magnetic resonance imaging(MRI) scans that most commonly reflect small vessel cerebrovascular disease. Increased WMH volume is associated with risk and progression of Alzheimer's disease(AD). These observations are typically interpreted as evidence that vascular abnormalities play an additive, independent role contributing to symptom presentation, but not core features of AD. We examined the severity and distribution of WMH in presymptomatic PSEN1, PSEN2, and APP mutation carriers to determine the extent to which WMH manifest in individuals genetically-determined to develop AD. Methods: The study comprised participants(n=299, age=39.03±10.13) from the Dominantly Inherited Alzheimer Network, including 184(61.5%) with a mutation that results in AD and 115(38.5%) first-degree relatives who were non-carrier controls. We calculated the estimated years from expected symptom onset(EYO) by subtracting the affected parent's symptom onset age from the participant's age. Baseline MRI data were analyzed for total and regional WMH. Mixed effects piecewise linear regression was used to examine WMH differences between carriers and non-carriers with respect to EYO. Results: Mutation carriers had greater total WMH volumes, which appeared to increase approximately 6 years prior to expected symptom onset. The effects were most prominent for the parietal and occipital lobe, which showed divergent effects as early as 22 years prior to estimated onset. Interpretation: Autosomal dominant AD is associated with increased WMH well before expected symptom onset. The findings suggest the possibility that WMH are a core feature of AD, a potential therapeutic target, and a factor that should be integrated into pathogenic models of the disease. This article is protected by copyright. All rights reserved.
Article
During the past decade, a conceptual shift occurred in the field of Alzheimer’s disease (AD) considering the disease as a continuum. Thanks to evolving biomarker research and substantial discoveries, it is now possible to identify the disease even at the preclinical stage before the occurrence of the first clinical symptoms. This preclinical stage of AD has become a major research focus as the field postulates that early intervention may offer the best chance of therapeutic success. To date, very little evidence is established on this “silent” stage of the disease. A clarification is needed about the definitions and lexicon, the limits, the natural history, the markers of progression, and the ethical consequence of detecting the disease at this asymptomatic stage. This article is aimed at addressing all the different issues by providing for each of them an updated review of the literature and evidence, with practical recommendations.
Article
Background: 1⁸F-FDFG uptake by brain tissue as measured by positron emission tomography (PET) is a well-established method for assessment of brain function in people with dementia. Certain findings on brain PET scans can potentially predict the decline of mild cognitive Impairment (MCI) to Alzheimer's disease dementia or other dementias. Objectives: To determine the diagnostic accuracy of the 1⁸F-FDG PET index test for detecting people with MCI at baseline who would clinically convert to Alzheimer's disease dementia or other forms of dementia at follow-up. Search methods: We searched the Cochrane Register of Diagnostic Test Accuracy Studies, MEDLINE, EMBASE, Science Citation Index, PsycINFO, BIOSIS previews, LILACS, MEDION, (Meta-analyses van Diagnostisch Onderzoek), DARE (Database of Abstracts of Reviews of Effects), HTA (Health Technology Assessment Database), ARIF (Aggressive Research Intelligence Facility) and C-EBLM (International Federation of Clinical Chemistry and Laboratory Medicine Committee for Evidence-based Laboratory Medicine) databases to January 2013. We checked the reference lists of any relevant studies and systematic reviews for additional studies. Selection criteria: We included studies that evaluated the diagnostic accuracy of 1⁸F-FDG PET to determine the conversion from MCI to Alzheimer's disease dementia or to other forms of dementia, i.e. any or all of vascular dementia, dementia with Lewy bodies, and fronto-temporal dementia. These studies necessarily employ delayed verification of conversion to dementia and are sometimes labelled as 'delayed verification cross-sectional studies'. Data collection and analysis: Two blinded review authors independently extracted data, resolving disagreement by discussion, with the option to involve a third review author as arbiter if necessary. We extracted and summarised graphically the data for two-by-two tables. We conducted exploratory analyses by plotting estimates of sensitivity and specificity from each study on forest plots and in receiver operating characteristic (ROC) space. When studies had mixed thresholds, we derived estimates of sensitivity and likelihood ratios at fixed values (lower quartile, median and upper quartile) of specificity from the hierarchical summary ROC (HSROC) models. Main results: We included 14 studies (421 participants) in the analysis. The sensitivities for conversion from MCI to Alzheimer's disease dementia were between 25% and 100% while the specificities were between 15% and 100%. From the summary ROC curve we fitted we estimated that the sensitivity was 76% (95% confidence interval (CI): 53.8 to 89.7) at the included study median specificity of 82%. This equates to a positive likelihood ratio of 4.03 (95% CI: 2.97 to 5.47), and a negative likelihood ratio of 0.34 (95% CI: 0.15 to 0.75). Three studies recruited participants from the same Alzheimer's Disease Neuroimaging Initiative (ADNI) cohort but only the largest ADNI study (Herholz 2011) is included in the meta-analysis. In order to demonstrate whether the choice of ADNI study or discriminating brain region (Chételat 2003) or reader assessment (Pardo 2010) make a difference to the pooled estimate, we performed five additional analyses. At the median specificity of 82%, the estimated sensitivity was between 74% and 76%. There was no impact on our findings. In addition to evaluating Alzheimer's disease dementia, five studies evaluated the accuracy of 1⁸F-FDG PET for all types of dementia. The sensitivities were between 46% and 95% while the specificities were between 29% and 100%; however, we did not conduct a meta-analysis because of too few studies, and those studies which we had found recruited small numbers of participants. Our findings are based on studies with poor reporting, and the majority of included studies had an unclear risk of bias, mainly for the reference standard and participant selection domains. According to the assessment of Index test domain, more than 50% of studies were of poor methodological quality. Authors' conclusions: It is difficult to determine to what extent the findings from the meta-analysis can be applied to clinical practice. Given the considerable variability of specificity values and lack of defined thresholds for determination of test positivity in the included studies, the current evidence does not support the routine use of 1⁸F-FDG PET scans in clinical practice in people with MCI. The 1⁸F-FDG PET scan is a high-cost investigation, and it is therefore important to clearly demonstrate its accuracy and to standardise the process of 1⁸F-FDG PET diagnostic modality prior to its being widely used. Future studies with more uniform approaches to thresholds, analysis and study conduct may provide a more homogeneous estimate than the one available from the included studies we have identified. © 2015 The Cochrane Collaboration. Published by John Wiley & Sons, Ltd.
Article
Brain imaging and fluid biomarkers are characterized in children at risk for autosomal dominant Alzheimer disease (ADAD). To characterize and compare structural magnetic resonance imaging (MRI), resting-state and task-dependent functional MRI, and plasma amyloid-β (Aβ) measurements in presenilin 1 (PSEN1) E280A mutation-carrying and noncarrying children with ADAD. Cross-sectional measures of structural and functional MRI and plasma Aβ assays were assessed in 18 PSEN1 E280A carriers and 19 noncarriers aged 9 to 17 years from a Colombian kindred with ADAD. Recruitment and data collection for this study were conducted at the University of Antioquia and the Hospital Pablo Tobon Uribe in Medellín, Colombia, between August 2011 and June 2012. All participants had blood sampling, structural MRI, and functional MRI during associative memory encoding and resting-state and cognitive assessments. Outcome measures included plasma Aβ1-42 concentrations and Aβ1-42:Aβ1-40 ratios, memory encoding-dependent activation changes, resting-state connectivity, and regional gray matter volumes. Structural and functional MRI data were compared using automated brain mapping algorithms and search regions related to AD. Similar to findings in adult mutation carriers, in the later preclinical and clinical stages of ADAD, mutation-carrying children were distinguished from control individuals by significantly higher plasma Aβ1-42 levels (mean [SD]: carriers, 18.8 [5.1] pg/mL and noncarriers, 13.1 [3.2] pg/mL; P < .001) and Aβ1-42:Aβ1-40 ratios (mean [SD]: carriers, 0.32 [0.06] and noncarriers, 0.21 [0.03]; P < .001), as well as less memory encoding task-related deactivation in parietal regions (eg, mean [SD] parameter estimates for the right precuneus were -0.590 [0.50] for noncarriers and -0.087 [0.38] for carriers; P < .005 uncorrected). Unlike carriers in the later stages, mutation-carrying children demonstrated increased functional connectivity of the posterior cingulate cortex with medial temporal lobe regions (mean [SD] parameter estimates were 0.038 [0.070] for noncarriers and 0.190 [0.057] for carriers), as well as greater gray matter volumes in temporal regions (eg, left parahippocampus; P < . 049, corrected for multiple comparisons). Children at genetic risk for ADAD have functional and structural brain changes and abnormal levels of plasma Aβ1-42. The extent to which the underlying brain changes are either neurodegenerative or developmental remains to be determined. This study provides additional information about the earliest known biomarker changes associated with ADAD.
Article
We present a technique for automatically assigning a neuroanatomical label to each location on a cortical surface model based on probabilistic information estimated from a manually labeled training set. This procedure incorporates both geometric information derived from the cortical model, and neuroanatomical convention, as found in the training set. The result is a complete labeling of cortical sulci and gyri. Examples are given from two different training sets generated using different neuroanatomical conventions, illustrating the flexibility of the algorithm. The technique is shown to be comparable in accuracy to manual labeling.
Article
Prior studies indicate psychiatric symptoms such as depression, apathy and anxiety are risk factors for or prodromal symptoms of incipient Alzheimer's disease. The study of persons at 50% risk for inheriting autosomal dominant Alzheimer's disease mutations allows characterization of these symptoms before progressive decline in a population destined to develop illness. We sought to characterize early behavioural features in carriers of autosomal dominant Alzheimer's disease mutations. Two hundred and sixty-one persons unaware of their mutation status enrolled in the Dominantly Inherited Alzheimer Network, a study of persons with or at-risk for autosomal dominant Alzheimer's disease, were evaluated with the Neuropsychiatric Inventory-Questionnaire, the 15-item Geriatric Depression Scale and the Clinical Dementia Rating Scale (CDR). Ninety-seven asymptomatic (CDR = 0), 25 mildly symptomatic (CDR = 0.5), and 33 overtly affected (CDR > 0.5) autosomal dominant Alzheimer's disease mutation carriers were compared to 106 non-carriers with regard to frequency of behavioural symptoms on the Neuropsychiatric Inventory-Questionnaire and severity of depressive symptoms on the Geriatric Depression Scale using generalized linear regression models with appropriate distributions and link functions. Results from the adjusted analyses indicated that depressive symptoms on the Neuropsychiatric Inventory-Questionnaire were less common in cognitively asymptomatic mutation carriers than in non-carriers (5% versus 17%, P = 0.014) and the odds of experiencing at least one behavioural sign in cognitively asymptomatic mutation carriers was lower than in non-carriers (odds ratio = 0.50, 95% confidence interval: 0.26-0.98, P = 0.042). Depression (56% versus 17%, P = 0.0003), apathy (40% versus 4%, P < 0.0001), disinhibition (16% versus 2%, P = 0.009), irritability (48% versus 9%, P = 0.0001), sleep changes (28% versus 7%, P = 0.003), and agitation (24% versus 6%, P = 0.008) were more common and the degree of self-rated depression more severe (mean Geriatric Depression Scale score of 2.8 versus 1.4, P = 0.006) in mildly symptomatic mutation carriers relative to non-carriers. Anxiety, appetite changes, delusions, and repetitive motor activity were additionally more common in overtly impaired mutation carriers. Similar to studies of late-onset Alzheimer's disease, we demonstrated increased rates of depression, apathy, and other behavioural symptoms in the mildly symptomatic, prodromal phase of autosomal dominant Alzheimer's disease that increased with disease severity. We did not identify any increased psychopathology in mutation carriers over non-carriers during the presymptomatic stage, suggesting these symptoms result when a threshold of neurodegeneration is reached rather than as life-long qualities. Unexpectedly, we found lower rates of depressive symptoms in cognitively asymptomatic mutation carriers. © The Author (2015). Published by Oxford University Press on behalf of the Guarantors of Brain. All rights reserved. For Permissions, please email: journals.permissions@oup.com.
Article
Search > Medical Terms (MeSH) > Search Manager > ARTICLE TOOLS Save to My Profile Export Citation for this Article E-mail Link to this Article Submit Comments More Articles like this Request Permissions More Sharing ServicesShare|Share on citeulikeShare on facebookShare on deliciousShare on www.mendeley.comShare on twitter Abstract Article Figures Tables References Other Versions Cited By View Full Article (HTML) Enhanced Article (HTML) Summary (112K)Standard (1429K)Full (1635K) Abstract Background ¹⁸F-FDFG uptake by brain tissue as measured by positron emission tomography (PET) is a well-established method for assessment of brain function in people with dementia. Certain findings on brain PET scans can potentially predict the decline of mild cognitive Impairment (MCI) to Alzheimer’s disease dementia or other dementias. Objectives To determine the diagnostic accuracy of the ¹⁸F-FDG PET index test for detecting people with MCI at baseline who would clinically convert to Alzheimer’s disease dementia or other forms of dementia at follow-up. Search methods We searched the Cochrane Register of Diagnostic Test Accuracy Studies, MEDLINE, EMBASE, Science Citation Index, PsycINFO, BIOSIS previews, LILACS, MEDION, (Meta-analyses van Diagnostisch Onderzoek), DARE (Database of Abstracts of Reviews of Effects), HTA (Health Technology Assessment Database), ARIF (Aggressive Research Intelligence Facility) and C-EBLM (International Federation of Clinical Chemistry and Laboratory Medicine Committee for Evidence-based Laboratory Medicine) databases to January 2013. We checked the reference lists of any relevant studies and systematic reviews for additional studies. Selection criteria We included studies that evaluated the diagnostic accuracy of ¹⁸F-FDG PET to determine the conversion from MCI to Alzheimer’s disease dementia or to other forms of dementia, i.e. any or all of vascular dementia, dementia with Lewy bodies, and fronto-temporal dementia. These studies necessarily employ delayed verification of conversion to dementia and are sometimes labelled as ‘delayed verification cross-sectional studies’. Data collection and analysis Two blinded review authors independently extracted data, resolving disagreement by discussion, with the option to involve a third review author as arbiter if necessary. We extracted and summarised graphically the data for two-by-two tables. We conducted exploratory analyses by plotting estimates of sensitivity and specificity from each study on forest plots and in receiver operating characteristic (ROC) space. When studies had mixed thresholds, we derived estimates of sensitivity and likelihood ratios at fixed values (lower quartile, median and upper quartile) of specificity from the hierarchical summary ROC (HSROC) models. Main results We included 14 studies (421 participants) in the analysis. The sensitivities for conversion from MCI to Alzheimer's disease dementia were between 25% and 100% while the specificities were between 15% and 100%. From the summary ROC curve we fitted we estimated that the sensitivity was 76% (95% confidence interval (CI): 53.8 to 89.7) at the included study median specificity of 82%. This equates to a positive likelihood ratio of 4.03 (95% CI: 2.97 to 5.47), and a negative likelihood ratio of 0.34 (95% CI: 0.15 to 0.75). Three studies recruited participants from the same Alzheimer's Disease Neuroimaging Initiative (ADNI) cohort but only the largest ADNI study (Herholz 2011) is included in the meta-analysis. In order to demonstrate whether the choice of ADNI study or discriminating brain region (Chételat 2003) or reader assessment (Pardo 2010) make a difference to the pooled estimate, we performed five additional analyses. At the median specificity of 82%, the estimated sensitivity was between 74% and 76%. There was no impact on our findings. In addition to evaluating Alzheimer's disease dementia, five studies evaluated the accuracy of ¹⁸F-FDG PET for all types of dementia. The sensitivities were between 46% and 95% while the specificities were between 29% and 100%; however, we did not conduct a meta-analysis because of too few studies, and those studies which we had found recruited small numbers of participants. Our findings are based on studies with poor reporting, and the majority of included studies had an unclear risk of bias, mainly for the reference standard and participant selection domains. According to the assessment of Index test domain, more than 50% of studies were of poor methodological quality. Authors' conclusions It is difficult to determine to what extent the findings from the meta-analysis can be applied to clinical practice. Given the considerable variability of specificity values and lack of defined thresholds for determination of test positivity in the included studies, the current evidence does not support the routine use of ¹⁸F-FDG PET scans in clinical practice in people with MCI. The ¹⁸F-FDG PET scan is a high-cost investigation, and it is therefore important to clearly demonstrate its accuracy and to standardise the process of ¹⁸F-FDG PET diagnostic modality prior to its being widely used. Future studies with more uniform approaches to thresholds, analysis and study conduct may provide a more homogeneous estimate than the one available from the included studies we have identified.
Article
Objective: To identify factors influencing age at symptom onset and disease course in autosomal dominant Alzheimer disease (ADAD), and develop evidence-based criteria for predicting symptom onset in ADAD. Methods: We have collected individual-level data on ages at symptom onset and death from 387 ADAD pedigrees, compiled from 137 peer-reviewed publications, the Dominantly Inherited Alzheimer Network (DIAN) database, and 2 large kindreds of Colombian (PSEN1 E280A) and Volga German (PSEN2 N141I) ancestry. Our combined dataset includes 3,275 individuals, of whom 1,307 were affected by ADAD with known age at symptom onset. We assessed the relative contributions of several factors in influencing age at onset, including parental age at onset, age at onset by mutation type and family, and APOE genotype and sex. We additionally performed survival analysis using data on symptom onset collected from 183 ADAD mutation carriers followed longitudinally in the DIAN Study. Results: We report summary statistics on age at onset and disease course for 174 ADAD mutations, and discover strong and highly significant (p < 10(-16), r2 > 0.38) correlations between individual age at symptom onset and predicted values based on parental age at onset and mean ages at onset by mutation type and family, which persist after controlling for APOE genotype and sex. Conclusions: Significant proportions of the observed variance in age at symptom onset in ADAD can be explained by family history and mutation type, providing empirical support for use of these data to estimate onset in clinical research.
Article
Background This study examined the predictive value of different classes of markers in the progression from mild cognitive impairment (MCI) to Alzheimer's disease (AD) over an extended 4-year follow-up in the Alzheimer's Disease Neuroimaging Initiative (ADNI) database. Methods MCI patients were assessed for clinical, cognitive, magnetic resonance imaging (MRI), positron emission tomography–fluorodeoxyglucose (PET-FDG), and cerebrospinal fluid (CSF) markers at baseline and were followed on a yearly basis for 4 years to ascertain progression to AD. Logistic regression models were fitted in clusters, including demographics, APOE genotype, cognitive markers, and biomarkers (morphometric, PET-FDG, CSF, amyloid-β, and tau). Results The predictive model at 4 years revealed that two cognitive measures, an episodic memory measure and a Clock Drawing screening test, were the best predictors of conversion (area under the curve = 0.78). Conclusions This model of prediction is consistent with the previous model at 2 years, thus highlighting the importance of cognitive measures in progression from MCI to AD. Cognitive markers were more robust predictors than biomarkers.
Article
In recent years, several approaches to develop computer aided diagnosis (CAD) systems for dementia have been proposed. Some of these systems analyze neurological brain images by means of machine learning algorithms in order to find the patterns that characterize the disorder, and a few combine several imaging modalities to improve the diagnostic accuracy. However, they usually do not use neuropsychological testing data in that analysis. The purpose of this work is to measure the advantages of using not only neuroimages as data source in CAD systems for dementia but also neuropsychological scores. To this aim, we compared the accuracy rates achieved by systems that use neuropsychological scores beside the imaging data in the classification step and systems that use only one of these data sources. In order to address the small sample size problem and facilitate the data combination, a dimensionality reduction step (implemented using three different algorithms) was also applied on the imaging data. After each image is summarized in a reduced set of image features, the data sources were combined and classified using three different data combination approaches and a Support Vector Machine classifier. That way, by testing different dimensionality reduction methods and several data combination approaches, we aim not only highlighting the advantages of using neuropsychological scores in the classification, but also implementing the most accurate computer system for early dementia detention. The accuracy of the CAD systems were estimated using a database with records from 46 subjects, diagnosed with MCI or AD. A peak accuracy rate of 89% was obtained. In all cases the accuracy achieved using both, neuropsychological scores and imaging data, was substantially higher than the one obtained using only the imaging data.
Article
Objective: To assess regional patterns of gray and white matter atrophy in familial Alzheimer disease (FAD) mutation carriers. Methods: A total of 192 participants with volumetric T1-weighted MRI, genotyping, and clinical diagnosis were available from the Dominantly Inherited Alzheimer Network. Of these, 69 were presymptomatic mutation carriers, 50 were symptomatic carriers (31 with Clinical Dementia Rating [CDR] = 0.5, 19 with CDR > 0.5), and 73 were noncarriers from the same families. Voxel-based morphometry was used to identify cross-sectional group differences in gray matter and white matter volume. Results: Significant differences in gray matter (p < 0.05, family-wise error-corrected) were observed between noncarriers and mildly symptomatic (CDR = 0.5) carriers in the thalamus and putamen, as well as in the temporal lobe, precuneus, and cingulate gyrus; the same pattern, but with more extensive changes, was seen in those with CDR > 0.5. Significant white matter differences between noncarriers and symptomatic carriers were observed in the cingulum and fornix; these form input and output connections to the medial temporal lobe, cingulate, and precuneus. No differences between noncarriers and presymptomatic carriers survived correction for multiple comparisons, but there was a trend for decreased gray matter in the thalamus for carriers closer to their estimated age at onset. There were no significant increases of gray or white matter in asymptomatic or symptomatic carriers compared to noncarriers. Conclusions: Atrophy in FAD is observed early, both in areas commonly associated with sporadic Alzheimer disease and also in the putamen and thalamus, 2 regions associated with early amyloid deposition in FAD mutation carriers.
Article
The Dominantly Inherited Alzheimer Network (DIAN) is an international registry of individuals at risk for developing autosomal dominant Alzheimer's disease (AD). Its primary aims are to investigate the temporal ordering of AD pathophysiological changes that occur in asymptomatic mutation carriers and to identify those markers that herald the transition from cognitive normality to symptomatic AD. DIAN participants undergo longitudinal evaluations, including clinical and cognitive assessments and measurements of molecular and imaging AD biomarkers. This review details the unique attributes of DIAN as a model AD biomarker study and how it provides the infrastructure for innovative research projects, including clinical trials. The recent design and launch of the first anti-amyloid-beta secondary prevention trial in AD, led by the related DIAN Trials Unit, also are discussed.
Article
The Dominantly Inherited Alzheimer's Network Trials Unit (DIAN-TU) was formed to direct the design and management of interventional therapeutic trials of international DIAN and autosomal dominant Alzheimer's disease (ADAD) participants. The goal of the DIAN-TU is to implement safe trials that have the highest likelihood of success while advancing scientific understanding of these diseases and clinical effects of proposed therapies. The DIAN-TU has launched a trial design that leverages the existing infrastructure of the ongoing DIAN observational study, takes advantage of a variety of drug targets, incorporates the latest results of biomarker and cognitive data collected during the observational study, and implements biomarkers measuring Alzheimer's disease (AD) biological processes to improve the efficiency of trial design. The DIAN-TU trial design is unique due to the sophisticated design of multiple drugs, multiple pharmaceutical partners, academics servings as sponsor, geographic distribution of a rare population and intensive safety and biomarker assessments. The implementation of the operational aspects such as home health research delivery, safety magnetic resonance imagings (MRIs) at remote locations, monitoring clinical and cognitive measures, and regulatory management involving multiple pharmaceutical sponsors of the complex DIAN-TU trial are described.
Article
We present a technique for automatically assigning a neuroanatomical label to each voxel in an MRI volume based on probabilistic information automatically estimated from a manually labeled training set. In contrast to existing segmentation procedures that only label a small number of tissue classes, the current method assigns one of 37 labels to each voxel, including left and right caudate, putamen, pallidum, thalamus, lateral ventricles, hippocampus, and amygdala. The classification technique employs a registration procedure that is robust to anatomical variability, including the ventricular enlargement typically associated with neurological diseases and aging. The technique is shown to be comparable in accuracy to manual labeling, and of sufficient sensitivity to robustly detect changes in the volume of noncortical structures that presage the onset of probable Alzheimer's disease.
Article
In the early days of PET, the partial volume effect (PVE) was identified as a serious factor affecting image quality and limiting the accuracy of quantitative analysis. Because of the limited spatial resolution of clinical PET systems, the images are blurred by the system response so that smaller objects appear larger. Although the total number of counts is preserved, they are distributed over a larger volume. This article describes the various partial volume correction strategies used in PET and summarizes their clinical and research applications.
Article
The article "Ten ironic rules for non-statistical reviewers" [4] shares some commonly heard frustrations about the peer-review process that all researchers can identify with. Though we found the article amusing, we have some concerns about its description of a number of statistical issues. In this commentary we address these issues, as well as the premise of the article.
Article
In 2010, we put forward a hypothetical model of the major biomarkers of Alzheimer's disease (AD). The model was received with interest because we described the temporal evolution of AD biomarkers in relation to each other and to the onset and progression of clinical symptoms. Since then, evidence has accumulated that supports the major assumptions of this model. Evidence has also appeared that challenges some of our assumptions, which has allowed us to modify our original model. Refinements to our model include indexing of individuals by time rather than clinical symptom severity; incorporation of interindividual variability in cognitive impairment associated with progression of AD pathophysiology; modifications of the specific temporal ordering of some biomarkers; and recognition that the two major proteinopathies underlying AD biomarker changes, amyloid β (Aβ) and tau, might be initiated independently in sporadic AD, in which we hypothesise that an incident Aβ pathophysiology can accelerate antecedent limbic and brainstem tauopathy.