Electrophysiological brain signatures for the classification of subjective cognitive decline: towards an individual detection in the preclinical stages of dementia

Abstract

Background Alzheimer’s disease (AD) prevalence is rapidly growing as worldwide populations grow older. Available treatments have failed to slow down disease progression, thus increasing research focus towards early or preclinical stages of the disease. Subjective cognitive decline (SCD) is known to increase the risk of developing AD and several other negative outcomes. However, it is still very scarcely characterized and there is no neurophysiological study devoted to its individual classification which could improve targeted sample recruitment for clinical trials. Methods Two hundred fifty-two older adults (70 healthy controls, 91 SCD, and 91 MCI) underwent a magnetoencephalography scan. Alpha relative power in the source space was employed to train a LASSO classifier and applied to distinguish between healthy controls and SCD. Moreover, MCI participants were used to further validate the previously trained algorithm. Results The classifier was significantly associated to SCD with an AUC of 0.81 in the whole sample. After randomly splitting the sample in 2/3 for discovery and 1/3 for validation, the newly trained classifier was also able to correctly classify SCD individuals with an AUC of 0.75 in the validation sample. The regions selected by the algorithm included medial frontal, temporal, and occipital areas. The algorithm trained to select SCD individuals was also significantly associated to MCI diagnostic. Conclusions According to our results, magnetoencephalography could be a useful tool for distinguishing individuals with SCD and healthy older adults without cognitive concerns. Furthermore, our classifier showed good external validity, being not only successful for an unseen SCD sample, but also in a different population with MCI cases. This supports its utility in the context of preclinical dementia. These findings highlight the potential applications of electrophysiological techniques to improve sample recruitment at the individual level in the context of clinical trials. Electronic supplementary material The online version of this article (10.1186/s13195-019-0502-3) contains supplementary material, which is available to authorized users.

Functional neuroimaging in subjective cognitive decline: current status and a research path forward

“…Multifaceted patterns of increased and decreased connectivity between regions could reflect breakdown in processing within a subnetwork that shifts processing load to more intact components of the broader network. Diffuse and frontally localized MEG alpha band power reductions have also been noted in older adults with SCD [39,46], without group differences in the small world properties of theta and beta bands [58]. Moreover, lower subgraph centrality (a measure of weighted, closed walks starting and ending at a node, representing mid-scale connectivity) has been observed in occipital and paracentral regions in older adults with SCD [48].…”

Gene Set Index Based on Different Modules May Help Differentiate the Mechanisms of Alzheimer’s Disease and Vascular Dementia

“…The least absolute shrinkage and selection operator (LASSO) is effective at selecting features from high-dimensional data that can be used to make classifications. 33 , 34 Firstly, GSE122063 data set was randomly assigned to the training set (70%) and test set (30%). We used the cv.glmnet function from glmnet package ( https://CRAN.R-project.org/package=glmnet ) to select the optimal genes with non-zero coefficients, which were entered into a LASSO model to differentiate among patients with AD, patients with vascular dementia and non-demented controls.…”

Identification of molecular correlations of RBM8A with autophagy in Alzheimer's disease

“…Least absolute shrinkage and selection operator (LASSO) has strong predictive value and low correlation and applied to select the best features for high-dimensional data [55, 56]. In order to distinguish AD from control, we extracted the expression profile of hub genes to construct LASSO model by glmnet package (https://CRAN.R-project.org/package=glmnet).…”