Monthly Archives: January 2022

Deep learning techniques applied to EEG data

Early yesterday morning I attended a remote conference talk, organized by ANT neuromeeting, and was fascinated by Dr. Sebastain Olbrich’s talk on his most recent paper on leveraging deep learning techniques applied to EEG data to help predict in conducting research. I go a bit into in on my YouTube (see below).

Curious to see if Dr. Olbrich would be interested for an interview on our channel. Will reach out to him and keep you posted!

In the meantime, here is the abstract of their paper:

Deep learning applied to electroencephalogram data in mental disorders: A systematic review

Mateo de Bardeci 1Cheng Teng Ip 2Sebastian Olbrich 3Affiliations expand

Free article


In recent medical research, tremendous progress has been made in the application of deep learning (DL) techniques. This article systematically reviews how DL techniques have been applied to electroencephalogram (EEG) data for diagnostic and predictive purposes in conducting research on mental disorders. EEG-studies on psychiatric diseases based on the ICD-10 or DSM-V classification that used either convolutional neural networks (CNNs) or long -short-term-memory (LSTMs) networks for classification were searched and examined for the quality of the information they contained in three domains: clinical, EEG-data processing, and deep learning. Although we found that the description of EEG acquisition and pre-processing was sufficient in most of the studies, we found, that many of them lacked a systematic characterization of clinical features. Furthermore, many studies used misguided model selection procedures or flawed testing. It is recommended that the study of psychiatric disorders using DL in the future must improve the quality of clinical data and follow state of the art model selection and testing procedures so as to achieve a higher research standard and head toward a clinical significance.

Keywords: CNN; Deep learning; Electroencephalogram; LSTM; Mental disorders.


Paperback launched: A Step-by-Step Guide to Collecting and Analyzing EEG Data with Emotiv EPOC Neuroheadset Series & EEGLAB MATLAB

A few people have been asking for the paperback version of the eBook because they felt that having this step-by-step guide in printed format would be more useful to them. The paperback edition is now available here: Paperback version (96 pages) for those interested in printed books.

The following is a summary of content of the book, followed by the table to contents, for your reference:

This book covers a step-by-step guide on the mechanics of setting up, collecting, processing, and visualizing raw EEG data (brainwaves) using Emotiv EPOC neuroheadset series with the EEGLAB (MATLAB) compiled environment open-source software. The step-by-step guide covered in this volume covers the basic mechanics and is not about complex or one-off EEG cases. This guide is ideal for people who want to get into EEG research, using Emotiv & EEGLAB, but do not know how to start. This guide should work well with all of the EPOC neuroheadset series; the EPOC, EPOC+, as well as EPOCx neuroheadsets. Whether you are a researcher, practitioner, or simply interested in the human brain, you will find it useful to study brainwaves primarily because brain frequencies tend to tell us how humans respond to stimuli at the neurological level. EEG data (brain frequencies or brainwaves) has several benefits compared to other imaging techniques or pure behavioral observations. This manual is the type of step-by-step guide that I wished I had when I was doing my grad studies. If you are interested in human brainwaves and want to learn one particular way of collecting and analyzing raw brain frequencies (EEG data), this volume is for you! By the end of this eBook, you will feel confident and uplifted enough to autonomously collect, analyze, and visualize raw brainwaves (EEG data). You will also feel confident to expand on the scope of this volume and showcase your EEG data analysis.

Here is also a summary of the content the book (Paperback version (96 pages)):


Chapter 1: Brainwaves

Overview: Electroencephalogram (EEG)

Overview: Brainwave types

Alpha Brainwaves (~ 8-12 HZ)

Beta Brainwaves (~ 13-30 HZ)

Gamma Brainwaves (~ 30+ HZ)

The international 10-20 EEG system

Chapter 2: Hardware

Setup: Emotiv EPOC Neuroheadset

Step 1: Charging and pairing the EPOC neuroheadset

Step 2: Hydrating the EPOC neuroheadset sensors

Step 3: Installing the EPOC neuroheadset sensors

Step 4: Fitting the EPOC neuroheadset

Step 5: EPOC neuroheadset placement and contact quality

Chapter 3: Software

Setup: EPOC control panel & EELAB MATLAB

Step 1: EPOC control panel software setup

Step 2: EEGLAB MATLAB setup

Chapter 4: Collecting raw EEG data

Section Five: Processing the raw EEG Data

Step 1: Starting EEGLAB MATLAB

Step 2: Importing the EEG data

Step 3: Loading raw EEG data

Step 4: Locating the neuroheadset channels

Creating CED file

Step 5: Reading the neuroheadset channel locations

Step 6: Running ICA and removing baseline

Step 7: Plotting to manually remove remaining artifacts

Chapter 6: Visualizing the EEG Data

Step 1: Channel spectra and maps

Step 2: Component maps in 2D

Step 3: Component maps in 3D

Section Seven: Interpreting EEG data


Visualizing the brain activities (EEG) during the Exploration stage

In the previous post, we covered the neural oscillations of the Exploration stage of the ISP model. However, the material was explained in either text or in table format. In this post, we will attempt to visualize the findings of this portion of the study instead.

To recap, the Exploration stage of the ISP model can be described:

Knowledge of a topic is gathered and a new personal knowledge is created. The individual endeavors to locate new knowledge to situate with previous understanding of the topic. Feelings of anxiety may be experienced if inconsistent and incompatible knowledge found.

One of the main focus of this study helped establish the neural oscillations of the ISP model. Simply put, we wanted to examine what happens in the brain during the Explorations stage.

As in, when we gather knowledge around a topic and when we situate that knowledge with previously accumulated information about the same topic, what happens to our brainwaves? Do they shift? If so, how do they change? And what if we felt either positive or negative when seeking knowledge? Would that change anything in our brain activities?

Note: This research controlled the neutral, positive, and negative conditions using the IAPS systems.

As explained in previous article, when the original state of being of the subjects was neutral, most of the human subjects showed Gamma bands (30+HZ) mainly in their left frontal lobes.

To illustrate, when feeling neutral while seeking (additional) information on a topic, our left frontal lobes show Gamma bands, which is 30+HZ in frequency:

But when the subjects emotional states were controlled towards positive or negative states, the neural oscillations as well as the locations of the bands actually changed.

In positive conditions, the left frontal lobe turned from Gamma to Beta bands. In other words, when feeling positive seeking (additional) information on a topic, our left frontal lobes show Beta bands (15-30HZ) frequency:

On the other hand, feeling negative when seeking (additional) information on a topic, we saw Gamma bands (30+HZ) in left frontal lobes AND right temporal and parietal lobs. Whereas Beta bands (15-30HZ) in the right frontal lobe.

What do these all mean though? In the next post, we will attempt to go into the potential meanings of these bands activations in the various brain lobes.

The Exploration stage of the ISP model

Almost 30 years later and the Information Search Process (ISP) model for knowledge remains as one of the most cited key theoretical frameworks in the discipline of Information Science. One of the ISP stages is the Exploration Stage. In this stage:

Knowledge of a topic is gathered and a new personal knowledge is created. The individual endeavors to locate new knowledge to situate with previous understanding of the topic. Feelings of anxiety may be experienced if inconsistent and incompatible knowledge found.

Most recent studies have emphasized the need for baseline studies to begin establishing neurophysiological data that help create foundations in understanding the underlying neurophysiology of behavior in information retrieval, such as information searching (Mostafa & Gwisdka, 2016).

This thesis attempted to establish a solid baseline for future researchers to understand the neurophysiology as well as the affective path of the information search process within the field of information science.

One of the goals of this thesis was to integrate the disciplines of neuroscience, information science, and cognitive psychology, while exploring possible connections among:

1. The affective states of the ISP model (Kuhlthau, 1991).
2. The corresponding emotional dimensions, valence versus arousal (Russell, 1980).
3. The corresponding neural oscillations, brain activities: alpha, beta, gamma (Sarraf, 2019).

This illustration summarizes some of the high-level integrated findings:


To help simplify and shorten this blog post, Russell’s valence-arousal correspondent axis is not included in the above illustration. It is, however, part of the thesis.

For your reference, here are brain lobes mentioned above:


What this all means is open to interpretations, some of which I have addressed in the thesis.