Abdullah

Abdullah

I am a PhD student working EEG channel selection using machine learning.

Sitemap

A list of all the posts and pages found on the site. For you robots out there is an XML version available for digesting as well.

Pages

Posts

Future Blog Post

less than 1 minute read

Published:

This post will show up by default. To disable scheduling of future posts, edit config.yml and set future: false.

Blog Post number 4

less than 1 minute read

Published:

This is a sample blog post. Lorem ipsum I can’t remember the rest of lorem ipsum and don’t have an internet connection right now. Testing testing testing this blog post. Blog posts are cool.

Blog Post number 3

less than 1 minute read

Published:

This is a sample blog post. Lorem ipsum I can’t remember the rest of lorem ipsum and don’t have an internet connection right now. Testing testing testing this blog post. Blog posts are cool.

Blog Post number 2

less than 1 minute read

Published:

This is a sample blog post. Lorem ipsum I can’t remember the rest of lorem ipsum and don’t have an internet connection right now. Testing testing testing this blog post. Blog posts are cool.

Blog Post number 1

less than 1 minute read

Published:

This is a sample blog post. Lorem ipsum I can’t remember the rest of lorem ipsum and don’t have an internet connection right now. Testing testing testing this blog post. Blog posts are cool.

portfolio

publications

Electroencephalogram Channel Selection using Deep Q-Network

Published in 2023 International Conference on Recent Advances in Electrical, Electronics & Digital Healthcare Technologies (REEDCON), 2023

In brain-computer interfaces, electroencephalogram channel selection picks the most informative channels. To speed up the model training and improve accuracy by selecting a small number of optimal channels. In this study, we trained an agent that automatically learned the policy to choose an optimal channel, from given EEG data, even without hand engineering. We frame the problem of EEG channel selection as a Markov decision process (MDP), offer a productive method for parameterizing it, and then apply deep reinforcement learning (DRL) to solve it. After the agent has been trained, it tries to learn a policy for channel selection that directs it to choose channels sequentially while leveraging EEG signals and previously selected tracks. The study also offers two reward systems for the DRL environment simulation and analyzes them in trials. This is the first work to look at a DRL model for EEG data interpretation, opening up a new field of study and highlighting DRL’s immense potential in the brain-computer interface.

Recommended citation: Abdullah, I. Faye and M. R. Islam, "Electroencephalogram Channel Selection using Deep Q-Network," 2023 International Conference on Recent Advances in Electrical, Electronics & Digital Healthcare Technologies (REEDCON), New Delhi, India, 2023, pp. 340-344, doi: 10.1109/REEDCON57544.2023.10151281.
Download Paper

Artificial Neural Networks Solutions for Solving Differential Equations: A Focus and Example for Flow of Viscoelastic Fluid with Microrotation

Published in Semarak Ilmu Publishing, 2023

Physics-informed neural networks (PINN) are an artificial neural network (ANN) approach for solving differential equations. PINN offers an alternative to classical numerical methods. The paper discusses the applications of PINN in various domains by highlighting the advantages, challenges, limitations, and some future directions. For example, PINN is implemented to solve the differential equations describing the Flow of Viscoelastic Fluid with Microrotation at a Horizontal Circular Cylinder Boundary Layer. The differential equations resulting from a nondimensionalization process of the governing equations and the associated boundary conditions are solved using PINN. The obtained results using PINN are discussed and compared to other state-of-the-art methods. Future research might aim to increase the precision and effectiveness of PINN models for solving differential equations, either by adding more physics-based restrictions or multi-scale methods to expand their capabilities. Additionally, investigating new application domains like linked multi-physics issues or real-time simulation situations may help to increase the reach and significance of PINN approaches.

Recommended citation: Abdullah, Ibrahima Faye, & Laila Amera Aziz. (2023). Artificial Neural Networks Solutions for Solving Differential Equations: A Focus and Example for Flow of Viscoelastic Fluid with Microrotation. Journal of Advanced Research in Fluid Mechanics and Thermal Sciences, 112(1), 76–83. https://doi.org/10.37934/arfmts.112.1.7683.
Download Paper

k-adaptEEGCS: Adaptive Threshold Based Automatic EEG Channel Selection

Published in IEEE Sensors Letters, 2024

Electroencephalography (EEG) channel selection is crucial for improving the accuracy and efficiency of EEG-based brain-computer interfaces (BCI) and cognitive state monitoring systems. This research identifies the most informative EEG channels that provide maximum discriminative power for specific tasks or applications. However, the availability of multiple electrodes can lead to data redundancy and increased computational complexity. In addition, selecting suboptimal channels may result in poor signal quality and reduced classification accuracy. A method called k-adaptEEGCS is proposed in this study to address these challenges. k-adaptEEGCS utilizes a similarity-metric-based approach to measure the similarity of EEG channels within each cluster and identify the best EEG channels using an adaptive threshold. The results show that k-adaptEEGCS improves classification accuracy and reduces channel selection time in specific EEG groups compared to using all EEG channels. Furthermore, the efficacy and superiority of k -adaptEEGCS are demonstrated through an analysis of BCI competition datasets; the average accuracy and channel reduction rate achieved is 93.09% and 67%.

Recommended citation: Abdullah, I. Faye, M. Tanveer and A. Vurity, " k -AdaptEEGCS: Adaptive Threshold Based Automatic EEG Channel Selection," in IEEE Sensors Letters, vol. 8, no. 10, pp. 1-4, Oct. 2024, Art no. 7501004, doi: 10.1109/LSENS.2024.3458996. keywords:
Download Paper

G-EEGCS: Graph-based optimum electroencephalogram channel selection

Published in Biomedical Signal Processing and Control, 2024

Electroencephalography (EEG) is commonly used to measure brain activity in clinical research. However, the abundance of channels in EEG data poses several challenges, including increased computational complexity, noise interference, and decreased efficiency in data analysis. A new graph-based method called G-EEGCS has been developed to address these issues for EEG channel selection. The graph-based optimum EEG channel selection (G-EEGCS) method constructs a directed network by establishing channel connections based on their pairwise correlations. Statistical features are used to determine similarity, and an adjacency matrix is created to represent the connectivity between the EEG channels. The influence of each channel on information flow within the network is assessed using the centrality measures. By calculating the shortest paths between all channel pairs, the algorithm quantifies the probability of each channel serving as a bridge between different parts of the graph. Channels with high centrality scores, indicating their significance in the information flow, are given priority during the selection process. An adaptive threshold is applied to optimize channel selection, ensuring that only channels that exceed the threshold, exhibit specific characteristics, and align with the overall network structure are retained. This adaptive thresholding mechanism enhances the robustness and flexibility of the G-EEGCS method, enabling personalized channel selection tailored to individual EEG datasets. The G-EEGCS approach offers a promising solution for EEG channel selection, improving the interpretability and efficiency of EEG-based studies and applications for researchers. It achieved an average accuracy of 0.9125, outperforming state-of-the-art methods. This method provides a valuable means of addressing the challenges posed by the multitude of channels in EEG data, ultimately contributing to more effective and insightful analysis in clinical research.

Recommended citation: Abdullah, Faye, I., Yusoff, M. Z., & Belhaouari, S. B. (2024). G-EEGCS: Graph-based optimum electroencephalogram channel selection. Biomedical Signal Processing and Control, 98, 106763.
Download Paper

talks

teaching

Visual Programming

Foundation Students, Universiti Teknologi PETRONAS, Department of Information Technology, 2021

Data Analytics

Undergraduate Course, Universiti Teknologi PETRONAS, Department of Applied Mathematics, 2021

  • Classification and Regression Algorithms