Computational algorithm associated with electroencephalography proves effective to diagnose Alzheimer’s

Using two low-cost techniques, researchers in Brazil differentiate patients with the disease from healthy subjects. Next steps include refining the approach to diagnose the disease in its early stages.

By André Julião  |  Agência FAPESP – A combination of two low-cost techniques has been shown to be an effective approach to diagnosing Alzheimer’s disease, a progressive neurodegenerative central nervous system disorder. Researchers at São Paulo State University (UNESP) and the National Space Research Institute (INPE) in Brazil have developed a software-driven method for analyzing electroencephalogram (EEG) data so as to distinguish between healthy subjects and individuals with Alzheimer’s. 

The EEG technique measures the electrical activity of the brain and is routinely used to diagnose conditions such as epilepsy, schizophrenia, sleep disorders, brain tumors, and Parkinson’s disease, among others. 

An article reporting the study is published in PLOS ONE.

“The technique used in this study was proposed during my PhD research. It maps physiological data to a complex network so that the dynamics can be analyzed in terms of the characteristics of the associated network. EEGs for Alzheimer’s patients are known to display a decrease in high-frequency components and an increase in low-frequency components compared with healthy subjects. We observed that mapping of the EEG data for subjects with distinct dynamics [healthy versus diseased] resulted in networks with similarly distinct topologies, validating the effectiveness of the technique,” said Andriana Campanharo, principal investigator for the study. Campanharo is a professor at UNESP’s Botucatu Institute of Biosciences (IBB).

Electroencephalography measures electrical activity (in volts) generated by the synchronized activity of thousands of neurons at sub-second timescales. Voltage fluctuations at the scalp electrodes are amplified and digitized for visualization. 

A complex network is a structure described by a set of vertices, edges, and some kind of interaction among them so that they can be analyzed computationally.

“One of the main advances achieved by the study is the use of EEG, which is low-cost, has high time resolution and is widely available, to obtain valuable information on the brain dynamics of Alzheimer’s patients,” said Aruane Mello Pineda, first author of the article. The study was conducted during her master’s research at IBB-UNESP.

The dataset was compiled by researchers at Florida State University in the United States and came from EEGs for 48 volunteers with ages ranging from 53 to 85. Half (24) were healthy subjects, and the rest were probable Alzheimer’s patients. 

Most affected brain regions

As well as classifying the volunteers in this way, the researchers investigated the brain regions most affected by Alzheimer’s using EEG data for 19 different scalp locations and mapped the corresponding complex networks. According to the authors, statistical analysis of these networks showed that the electrical signals associated with the disease were best detected in the left temporal-parietal area.

“This is in line with the current understanding of Alzheimer’s, which mainly affects the left side of the temporal-hippocampal network, responsible for verbal memory and apparently the more vulnerable hemisphere,” said Pineda, now doing PhD research at the Center for Research in Mathematical Sciences Applied to Industry (CeMEAI), a Research, Innovation and Dissemination Center (RIDC) funded by FAPESP and hosted by the University of São Paulo (USP) in São Carlos.

The group led by Campanharo is working with a larger dataset covering more than 100 subjects who are healthy or have different stages of Alzheimer’s. Now that the technique has been validated to identify patients with advanced Alzheimer’s, the goal is to identify patterns that distinguish between the initial and advanced stages. This could lead to automatic and more precise diagnosis, with the possibility of detecting the disease sooner and facilitating early treatment.

Fernando M. Ramos, at INPE, and Luiz Eduardo Betting, at the Botucatu Medical School (FMB-UNESP), also collaborated on the study.

The article “Quantile graphs for EEG-based diagnosis of Alzheimer’s disease” is at: journals.plos.org/plosone/article?id=10.1371/journal.pone.0231169.