The PredictAD project: development of novel biomarkers and analysis software for early diagnosis of the Alzheimer's disease

Artificial intelligence for dementia research methods optimization

Artificial intelligence (AI) and machine learning (ML) approaches are increasingly being used in dementia research. However, several methodological challenges exist that may limit the insights we can obtain from high-dimensional data and our ability to translate these findings into improved patient outcomes. To improve reproducibility and replicability, researchers should make their well-documented code and modeling pipelines openly available. Data should also be shared where appropriate. To enhance the acceptability of models and AI-enabled systems to users, researchers should prioritize interpretable methods that provide insights into how decisions are generated. Models should be developed using multiple, diverse datasets to improve robustness, generalizability, and reduce potentially harmful bias. To improve clarity and reproducibility, researchers should adhere to reporting guidelines that are co-produced with multiple stakeholders. If these methodological challenges are overcome, AI and ML hold enormous promise for changing the landscape of dementia research and care. HIGHLIGHTS: Machine learning (ML) can improve diagnosis, prevention, and management of dementia. Inadequate reporting of ML procedures affects reproduction/replication of results. ML models built on unrepresentative datasets do not generalize to new datasets. Obligatory metrics for certain model structures and use cases have not been defined. Interpretability and trust in ML predictions are barriers to clinical translation.

Artificial Intelligence for Dementia Research Methods Optimization

INTRODUCTION

Machine learning (ML) has been extremely successful in identifying key features from high-dimensional datasets and executing complicated tasks with human expert levels of accuracy or greater.

METHODS

We summarize and critically evaluate current applications of ML in dementia research and highlight directions for future research.

RESULTS

We present an overview of ML algorithms most frequently used in dementia research and highlight future opportunities for the use of ML in clinical practice, experimental medicine, and clinical trials. We discuss issues of reproducibility, replicability and interpretability and how these impact the clinical applicability of dementia research. Finally, we give examples of how state-of-the-art methods, such as transfer learning, multi-task learning, and reinforcement learning, may be applied to overcome these issues and aid the translation of research to clinical practice in the future.

DISCUSSION

ML-based models hold great promise to advance our understanding of the underlying causes and pathological mechanisms of dementia.

Developing and validating a multivariable prediction model which predicts progression of intermediate to late age-related macular degeneration-the PINNACLE trial protocol

AIMS

Age-related macular degeneration (AMD) is characterised by a progressive loss of central vision. Intermediate AMD is a risk factor for progression to advanced stages categorised as geographic atrophy (GA) and neovascular AMD. However, rates of progression to advanced stages vary between individuals. Recent advances in imaging and computing technologies have enabled deep phenotyping of intermediate AMD. The aim of this project is to utilise machine learning (ML) and advanced statistical modelling as an innovative approach to discover novel features and accurately quantify markers of pathological retinal ageing that can individualise progression to advanced AMD.

METHODS

The PINNACLE study consists of both retrospective and prospective parts. In the retrospective part, more than 400,000 optical coherent tomography (OCT) images collected from four University Teaching Hospitals and the UK Biobank Population Study are being pooled, centrally stored and pre-processed. With this large dataset featuring eyes with AMD at various stages and healthy controls, we aim to identify imaging biomarkers for disease progression for intermediate AMD via supervised and unsupervised ML. The prospective study part will firstly characterise the progression of intermediate AMD in patients followed between one and three years; secondly, it will validate the utility of biomarkers identified in the retrospective cohort as predictors of progression towards late AMD. Patients aged 55-90 years old with intermediate AMD in at least one eye will be recruited across multiple sites in UK, Austria and Switzerland for visual function tests, multimodal retinal imaging and genotyping. Imaging will be repeated every four months to identify early focal signs of deterioration on spectral-domain optical coherence tomography (OCT) by human graders. A focal event triggers more frequent follow-up with visual function and imaging tests. The primary outcome is the sensitivity and specificity of the OCT imaging biomarkers. Secondary outcomes include sensitivity and specificity of novel multimodal imaging characteristics at predicting disease progression, ROC curves, time from development of imaging change to development of these endpoints, structure-function correlations, structure-genotype correlation and predictive risk models.

CONCLUSIONS

This is one of the first studies in intermediate AMD to combine both ML, retrospective and prospective AMD patient data with the goal of identifying biomarkers of progression and to report the natural history of progression of intermediate AMD with multimodal retinal imaging.

Induction of specific brain oscillations may restore neural circuits and be used for the treatment of Alzheimer's disease

Brain oscillations underlie the function of our brains, dictating how we both think and react to the world around us. The synchronous activity of neurons generates these rhythms, which allow different parts of the brain to communicate and orchestrate responses to internal and external stimuli. Perturbations of cognitive rhythms and the underlying oscillator neurons that synchronize different parts of the brain contribute to the pathophysiology of diseases including Alzheimer's disease, (AD), Parkinson's disease (PD), epilepsy and other diseases of rhythm that have been studied extensively by Gyorgy Buzsaki. In this review, we discuss how neurologists manipulate brain oscillations with neuromodulation to treat diseases and how this can be leveraged to improve cognition and pathology underlying AD. While multiple modalities of neuromodulation are currently clinically indicated for some disorders, nothing is yet approved for improving memory in AD. Recent investigations into novel methods of neuromodulation show potential for improving cognition in memory disorders. Here, we demonstrate that neuronal stimulation using audiovisual sensory stimulation that generated 40-HZ gamma waves reduced AD-specific pathology and improved performance in behavioural tests in mouse models of AD, making this new mode of neuromodulation a promising new avenue for developing a new therapeutic intervention for the treatment of dementia.

Current advances in digital cognitive assessment for preclinical Alzheimer's disease

There is a pressing need to capture and track subtle cognitive change at the preclinical stage of Alzheimer's disease (AD) rapidly, cost-effectively, and with high sensitivity. Concurrently, the landscape of digital cognitive assessment is rapidly evolving as technology advances, older adult tech-adoption increases, and external events (i.e., COVID-19) necessitate remote digital assessment. Here, we provide a snapshot review of the current state of digital cognitive assessment for preclinical AD including different device platforms/assessment approaches, levels of validation, and implementation challenges. We focus on articles, grants, and recent conference proceedings specifically querying the relationship between digital cognitive assessments and established biomarkers for preclinical AD (e.g., amyloid beta and tau) in clinically normal (CN) individuals. Several digital assessments were identified across platforms (e.g., digital pens, smartphones). Digital assessments varied by intended setting (e.g., remote vs. in-clinic), level of supervision (e.g., self vs. supervised), and device origin (personal vs. study-provided). At least 11 publications characterize digital cognitive assessment against AD biomarkers among CN. First available data demonstrate promising validity of this approach against both conventional assessment methods (moderate to large effect sizes) and relevant biomarkers (predominantly weak to moderate effect sizes). We discuss levels of validation and issues relating to usability, data quality, data protection, and attrition. While still in its infancy, digital cognitive assessment, especially when administered remotely, will undoubtedly play a major future role in screening for and tracking preclinical AD.

Effects of sex and estrous cycle on the brain and plasma arginine metabolic profile in rats

L-arginine is a versatile amino acid with a number of bioactive metabolites. Increasing evidence implicates altered arginine metabolism in the aging and neurodegenerative processes. The present study, for the first time, determined the effects of sex and estrous cycle on the brain and blood (plasma) arginine metabolic profile in naïve rats. Female rats displayed significantly lower levels of L-arginine in the frontal cortex and three sub-regions of the hippocampus when compared to male rats. Moreover, female rats had significantly higher levels of L-arginine and γ-aminobutyric acid, but lower levels of L-ornithine, agmatine and putrescine, in plasma relative to male rats. The observed sex difference in brain L-arginine appeared to be independent of the enzymes involved in its metabolism, de novo synthesis and blood-to-brain transport (cationic acid transporter 1 protein expression at least), as well as circulating L-arginine. While the estrous cycle did not affect L-arginine and its metabolites in the brain, there were estrous cycle phase-dependent changes in plasma L-arginine. These findings demonstrate the sex difference in brain L-arginine in the estrous cycle-independent manner. Since peripheral blood has been increasingly used to identify biomarkers of brain pathology, the influences of sex and estrous cycle on blood arginine metabolic profile need attention when experimental research involves female rodents.

Hypersynchronization in mild cognitive impairment: the ‘X’ model

Hypersynchronization has been considered as a biomarker of synaptic dysfunction along the Alzheimeŕs disease continuum. In a longitudinal MEG study, Pusil et al. reveal changes in functional connectivity upon progression from MCI to Alzheimer’s disease. They propose the ‘X’ model to explain their findings, and suggest that hypersynchronization predicts conversion.

A practical computerized decision support system for predicting the severity of Alzheimer's disease of an individual

Computerized clinical decision support systems can help to provide objective, standardized, and timely dementia diagnosis. However, current computerized systems are mainly based on group analysis, discrete classification of disease stages, or expensive and not readily accessible biomarkers, while current clinical practice relies relatively heavily on cognitive and functional assessments (CFA). In this study, we developed a computational framework using a suite of machine learning tools for identifying key markers in predicting the severity of Alzheimer's disease (AD) from a large set of biological and clinical measures. Six machine learning approaches, namely Kernel Ridge Regression (KRR), Support Vector Regression, and k-Nearest Neighbor for regression and Support Vector Machine (SVM), Random Forest, and k-Nearest Neighbor for classification, were used for the development of predictive models. We demonstrated high predictive power of CFA. Predictive performance of models incorporating CFA was shown to consistently have higher accuracy than those based solely on biomarker modalities. We found that KRR and SVM were the best performing regression and classification methods respectively. The optimal SVM performance was observed for a set of four CFA test scores (FAQ, ADAS13, MoCA, MMSE) with multi-class classification accuracy of 83.0%, 95%CI = (72.1%, 93.8%) while the best performance of the KRR model was reported with combined CFA and MRI neuroimaging data, i.e., R ² = 0.874, 95%CI = (0.827, 0.922). Given the high predictive power of CFA and their widespread use in clinical practice, we then designed a data-driven and self-adaptive computerized clinical decision support system (CDSS) prototype for evaluating the severity of AD of an individual on a continuous spectrum. The system implemented an automated computational approach for data pre-processing, modelling, and validation and used exclusively the scores of selected cognitive measures as data entries. Taken together, we have developed an objective and practical CDSS to aid AD diagnosis.

Serum Metabolites Associated with Computed Tomography Findings after Traumatic Brain Injury

There is a need to rapidly detect patients with traumatic brain injury (TBI) who require head computed tomography (CT). Given the energy crisis in the brain following TBI, we hypothesized that serum metabolomics would be a useful tool for developing a set of biomarkers to determine the need for CT and to distinguish among different types of injuries observed. Logistical regression models using metabolite data from the discovery cohort (n = 144, Turku, Finland) were used to distinguish between patients with traumatic intracranial findings and those with negative findings on head CT. The resultant models were then tested in the validation cohort (n = 66, Cambridge, United Kingdom). The levels of glial fibrillary acidic protein and ubiquitin C-terminal hydrolase-L1 were also quantified in the serum from the same patients. Despite there being significant differences in the protein biomarkers in patients with TBI, the model that determined the need for a CT scan validated poorly (area under the curve [AUC] = 0.64: Cambridge patients). However, using a combination of six metabolites (two amino acids, three sugar derivatives, and one ketoacid) it was possible to discriminate patients with intracranial abnormalities on CT and patients with a normal CT (AUC = 0.77 in Turku patients and AUC = 0.73 in Cambridge patients). Further, a combination of three metabolites could distinguish between diffuse brain injuries and mass lesions (AUC = 0.87 in Turku patients and AUC = 0.68 in Cambridge patients). This study identifies a set of validated serum polar metabolites, which associate with the need for a CT scan. Additionally, serum metabolites can also predict the nature of the brain injury. These metabolite markers may prevent unnecessary CT scans, thus reducing the cost of diagnostics and radiation load.

Altered plasma arginine metabolome precedes behavioural and brain arginine metabolomic profile changes in the APPswe/PS1ΔE9 mouse model of Alzheimer's disease

While amyloid-beta (Aβ) peptides play a central role in the development of Alzheimer's disease (AD), recent evidence also implicates altered metabolism of L-arginine in the pathogenesis of AD. The present study systematically investigated how behavioural function and the brain and plasma arginine metabolic profiles changed in a chronic Aβ accumulation model using male APPswe/PS1ΔE9 transgenic (Tg) mice at 7 and 13 months of age. As compared to their wild-type (WT) littermates, Tg mice displayed age-related deficits in spatial water maze tasks and alterations in brain arginine metabolism. Interestingly, the plasma arginine metabolic profile was markedly altered in 7-month Tg mice prior to major behavioural impairment. Receiver operating characteristic curve analysis revealed that plasma putrescine and spermine significantly differentiated between Tg and WT mice. These results demonstrate the parallel development of altered brain arginine metabolism and behavioural deficits in Tg mice. The altered plasma arginine metabolic profile that preceded the behavioural and brain profile changes suggests that there may be merit in an arginine-centric set of ante-mortem biomarkers for AD.

Interpreting Biomarker Results in Individual Patients With Mild Cognitive Impairment in the Alzheimer's Biomarkers in Daily Practice (ABIDE) Project

Importance

Biomarkers do not determine conversion to Alzheimer disease (AD) perfectly, and criteria do not specify how to take patient characteristics into account. Consequently, biomarker use may be challenging for clinicians, especially in patients with mild cognitive impairment (MCI).

Objective

To construct biomarker-based prognostic models that enable determination of future AD dementia in patients with MCI.

Design, Setting, and Participants

This study is part of the Alzheimer's Biomarkers in Daily Practice (ABIDE) project. A total of 525 patients with MCI from the Amsterdam Dementia Cohort (longitudinal cohort, tertiary referral center) were studied. All patients had their baseline visit to a memory clinic from September 1, 1997, through August 31, 2014. Prognostic models were constructed by Cox proportional hazards regression with patient characteristics (age, sex, and Mini-Mental State Examination [MMSE] score), magnetic resonance imaging (MRI) biomarkers (hippocampal volume, normalized whole-brain volume), cerebrospinal fluid (CSF) biomarkers (amyloid-β1-42, tau), and combined biomarkers. Data were analyzed from November 1, 2015, to October 1, 2016.

Main Outcomes and Measures

Clinical end points were AD dementia and any type of dementia after 1 and 3 years.

Results

Of the 525 patients, 210 (40.0%) were female, and the mean (SD) age was 67.3 (8.4) years. On the basis of age, sex, and MMSE score only, the 3-year progression risk to AD dementia ranged from 26% (95% CI, 19%-34%) in younger men with MMSE scores of 29 to 76% (95% CI, 65%-84%) in older women with MMSE scores of 24 (1-year risk: 6% [95% CI, 4%-9%] to 24% [95% CI, 18%-32%]). Three- and 1-year progression risks were 86% (95% CI, 71%-95%) and 27% (95% CI, 17%-41%) when MRI results were abnormal, 82% (95% CI, 73%-89%) and 26% (95% CI, 20%-33%) when CSF test results were abnormal, and 89% (95% CI, 79%-95%) and 26% (95% CI, 18%-36%) when the results of both tests were abnormal. Conversely, 3- and 1-year progression risks were 18% (95% CI, 13%-27%) and 3% (95% CI, 2%-5%) after normal MRI results, 6% (95% CI, 3%-9%) and 1% (95% CI, 0.5%-2%) after normal CSF test results, and 4% (95% CI, 2%-7%) and 0.5% (95% CI, 0.2%-1%) after combined normal MRI and CSF test results. The prognostic value of models determining any type of dementia were in the same order of magnitude although somewhat lower. External validation in Alzheimer's Disease Neuroimaging Initiative 2 showed that our models were highly robust.

Conclusions and Relevance

This study provides biomarker-based prognostic models that may help determine AD dementia and any type of dementia in patients with MCI at the individual level. This finding supports clinical decision making and application of biomarkers in daily practice.

Early Diagnosis of Alzheimer's Disease by Joint Feature Selection and Classification on Temporally Structured Support Vector Machine

The diagnosis of Alzheimer's disease (AD) from neuroimaging data at the pre-clinical stage has been intensively investigated because of the immense social and economic cost. In the past decade, computational approaches on longitudinal image sequences have been actively investigated with special attention to Mild Cognitive Impairment (MCI), which is an intermediate stage between normal control (NC) and AD. However, current state-of-the-art diagnosis methods have limited power in clinical practice, due to the excessive requirements such as equal and immoderate number of scans in longitudinal imaging data. More critically, very few methods are specifically designed for the early alarm of AD uptake. To address these limitations, we propose a flexible spatial-temporal solution for early detection of AD by recognizing abnormal structure changes from longitudinal MR image sequence. Specifically, our method is leveraged by the non-reversible nature of AD progression. We employ temporally structured SVM to accurately alarm AD at early stage by enforcing the monotony on classification result to avoid unrealistic and inconsistent diagnosis result along time. Furthermore, in order to select best features which can well collaborate with the classifier, we present as joint feature selection and classification framework. The evaluation on more than 150 longitudinal subjects from ADNI dataset shows that our method is able to alarm the conversion of AD 12 months prior to the clinical diagnosis with at least 82.5 % accuracy. It is worth noting that our proposed method works on widely used MR images and does not have restriction on the number of scans in the longitudinal sequence, which is very attractive to real clinical practice.

The relationship between baseline EEG spectra power and memory performance in older African Americans endorsing cognitive concerns in a community setting

The finding that some older individuals report declines in aspects of cognitive functioning is becoming a frequently used criteria to identify elderly at risk for mild cognitive impairment (MCI) and dementia. Once concerns are identified in a community setting, however, effective means are necessary to pinpoint those individuals who should go on to more complex and costly diagnostic evaluations (e.g., functional imaging). We tested 44 African American volunteers endorsing cognitive concerns (37 females, 7 males) age≥65years with CogState battery subtests and recorded resting-state EEG, with eyes closed. After current source density (CSD) transformations of EEG recordings we obtained spectral power for delta, theta, alpha, and beta frequency bands. We characterized CogState One Card Back Learning (OCL, memory) with diffusion model parameters drift rate, boundary and non-decision time (NDT). Forward regression models showed that lower OCL drift rate, slower accumulation of information needed for decision making was linked to increased absolute and relative delta at occipital region. Lower drift rate was also linked to decrease in OCL theta power at parietal region, with no findings for ONB. Results show that cortical resting, eyes closed EEG rhythms are related to memory in African American seniors endorsing cognitive concerns. This study further supports the use of EEG as an easily accessible, cost-effective, culture-fair, and noninvasive clinical measurement that could provide potentially reliable diagnostic (and perhaps prognostic) information to differentiate at-risk from stable African American seniors.

Quantitative-profiling of neurotransmitter abnormalities in the disease progression of experimental diabetic encephalopathy rat

Diabetic encephalopathy (DE) is one of the most prevalent chronic complications of diabetes mellitus (DM), with neither effective prevention nor proven therapeutic regimen. This study aims to uncover the potential dysregulation pattern of the neurotransmitters in a rat model of streptozotocin (STZ)-induced experimental DE. For that purpose, male Sprague–Dawley (SD) rats were treated with a single intraperitoneal injection of STZ. Cognitive performance was detected with the Morris water maze (MWM) test. Serum, cerebrospinal fluid (CSF), and brain tissues were collected to measure the levels of neurotransmitters. Compared with the control rats, the acetylcholine (ACh) levels in serum, CSF, hippocampus, and cortex were all significantly down-regulated as early as 6 weeks in the STZ treatment group. In contrast, the glutamate (Glu) levels were decreased in CSF and the hippocampus, but unaffected in the serum and cortex of STZ-treated rats. As for γ-aminobutyric acid (GABA), it was down-regulated in serum, but up-regulated in CSF, hippocampus, and the cortex in the STZ-treated group. The mRNA expressions of neurotransmitter-related rate limiting enzymes (including AChE, GAD1, and GAD2) and pro-inflammatory cytokines (including IL-1β and TNF-α) were all increased in the DE rats. Our data suggest that DM induces isoform-dependent and tissue-specific neurotransmitter abnormalities, and that neuroinflammation may underlay the nervous system dysfunction observed in the progression of DE.

MicroRNA-Seq Data Analysis Pipeline to Identify Blood Biomarkers for Alzheimer's Disease from Public Data

BACKGROUND

Alzheimer’s disease (AD) is the most common cause of dementia with no curative therapy currently available. Establishment of sensitive and non-invasive biomarkers that promote an early diagnosis of AD is crucial for the effective administration of disease-modifying drugs. MicroRNAs (miRNAs) mediate posttranscriptional repression of numerous target genes. Aberrant regulation of miRNA expression is implicated in AD pathogenesis, and circulating miRNAs serve as potential biomarkers for AD. However, data analysis of numerous AD-specific miRNAs derived from small RNA-sequencing (RNA-Seq) is most often laborious.

METHODS

To identify circulating miRNA biomarkers for AD, we reanalyzed a publicly available small RNA-Seq dataset, composed of blood samples derived from 48 AD patients and 22 normal control (NC) subjects, by a simple web-based miRNA data analysis pipeline that combines omiRas and DIANA miRPath.

RESULTS

By using omiRas, we identified 27 miRNAs expressed differentially between both groups, including upregulation in AD of miR-26b-3p, miR-28–3p, miR-30c-5p, miR-30d-5p, miR-148b-5p, miR-151a-3p, miR-186–5p, miR-425–5p, miR-550a-5p, miR-1468, miR-4781–3p, miR-5001–3p, and miR-6513–3p and downregulation in AD of let-7a-5p, let-7e-5p, let-7f-5p, let-7g-5p, miR-15a-5p, miR-17–3p, miR-29b-3p, miR-98–5p, miR-144–5p, miR-148a-3p, miR-502–3p, miR-660–5p, miR-1294, and miR-3200–3p. DIANA miRPath indicated that miRNA-regulated pathways potentially downregulated in AD are linked with neuronal synaptic functions, while those upregulated in AD are implicated in cell survival and cellular communication.

CONCLUSIONS

The simple web-based miRNA data analysis pipeline helps us to effortlessly identify candidates for miRNA biomarkers and pathways of AD from the complex small RNA-Seq data.

Alpha-band hypersynchronization in progressive mild cognitive impairment: a magnetoencephalography study

People with mild cognitive impairment (MCI) show a high risk to develop Alzheimer's disease (AD; Petersen et al., 2001). Nonetheless, there is a lack of studies about how functional connectivity patterns may distinguish between progressive (pMCI) and stable (sMCI) MCI patients. To examine whether there were differences in functional connectivity between groups, MEG eyes-closed recordings from 30 sMCI and 19 pMCI subjects were compared. The average conversion time of pMCI was 1 year, so they were considered as fast converters. To this end, functional connectivity in different frequency bands was assessed with phase locking value in source space. Then the significant differences between both groups were correlated with neuropsychological scores and entorhinal, parahippocampal, and hippocampal volumes. Both groups did not differ in age, gender, or educational level. pMCI patients obtained lower scores in episodic and semantic memory and also in executive functioning. At the structural level, there were no differences in hippocampal volume, although some were found in left entorhinal volume between both groups. Additionally, pMCI patients exhibit a higher synchronization in the alpha band between the right anterior cingulate and temporo-occipital regions than sMCI subjects. This hypersynchronization was inversely correlated with cognitive performance, both hippocampal volumes, and left entorhinal volume. The increase in phase synchronization between the right anterior cingulate and temporo-occipital areas may be predictive of conversion from MCI to AD.

Brain investigation and brain conceptualization

The brain of a patient with Alzheimer's disease (AD) undergoes changes starting many years before the development of the first clinical symptoms. The recent availability of large prospective datasets makes it possible to create sophisticated brain models of healthy subjects and patients with AD, showing pathophysiological changes occurring over time. However, these models are still inadequate; representations are mainly single-scale and they do not account for the complexity and interdependence of brain changes. Brain changes in AD patients occur at different levels and for different reasons: at the molecular level, changes are due to amyloid deposition; at cellular level, to loss of neuron synapses, and at tissue level, to connectivity disruption. All cause extensive atrophy of the whole brain organ. Initiatives aiming to model the whole human brain have been launched in Europe and the US with the goal of reducing the burden of brain diseases. In this work, we describe a new approach to earlier diagnosis based on a multimodal and multiscale brain concept, built upon existing and well-characterized single modalities.

Integrative EEG biomarkers predict progression to Alzheimer's disease at the MCI stage

Alzheimer's disease (AD) is a devastating disorder of increasing prevalence in modern society. Mild cognitive impairment (MCI) is considered a transitional stage between normal aging and AD; however, not all subjects with MCI progress to AD. Prediction of conversion to AD at an early stage would enable an earlier, and potentially more effective, treatment of AD. Electroencephalography (EEG) biomarkers would provide a non-invasive and relatively cheap screening tool to predict conversion to AD; however, traditional EEG biomarkers have not been considered accurate enough to be useful in clinical practice. Here, we aim to combine the information from multiple EEG biomarkers into a diagnostic classification index in order to improve the accuracy of predicting conversion from MCI to AD within a 2-year period. We followed 86 patients initially diagnosed with MCI for 2 years during which 25 patients converted to AD. We show that multiple EEG biomarkers mainly related to activity in the beta-frequency range (13–30 Hz) can predict conversion from MCI to AD. Importantly, by integrating six EEG biomarkers into a diagnostic index using logistic regression the prediction improved compared with the classification using the individual biomarkers, with a sensitivity of 88% and specificity of 82%, compared with a sensitivity of 64% and specificity of 62% of the best individual biomarker in this index. In order to identify this diagnostic index we developed a data mining approach implemented in the Neurophysiological Biomarker Toolbox (http://www.nbtwiki.net/). We suggest that this approach can be used to identify optimal combinations of biomarkers (integrative biomarkers) also in other modalities. Potentially, these integrative biomarkers could be more sensitive to disease progression and response to therapeutic intervention.

Brain investigation and brain conceptualization

The brain of a patient with Alzheimer's disease (AD) undergoes changes starting many years before the development of the first clinical symptoms. The recent availability of large prospective datasets makes it possible to create sophisticated brain models of healthy subjects and patients with AD, showing pathophysiological changes occurring over time. However, these models are still inadequate; representations are mainly single-scale and they do not account for the complexity and interdependence of brain changes. Brain changes in AD patients occur at different levels and for different reasons: at the molecular level, changes are due to amyloid deposition; at cellular level, to loss of neuron synapses, and at tissue level, to connectivity disruption. All cause extensive atrophy of the whole brain organ. Initiatives aiming to model the whole human brain have been launched in Europe and the US with the goal of reducing the burden of brain diseases. In this work, we describe a new approach to earlier diagnosis based on a multimodal and multiscale brain concept, built upon existing and well-characterized single modalities.