My Research Projects

(This research is currently conducted) - Stay tuned for interesting result soon!.
This research investigates the neural information processing mechanism and dynamics of the brain during the emergence of congnitive functions using computational and mathematical analysis of fMRI data

Hudgkin-Huxley Model are important tool in neuronal modelling, it captures the detailed gating properties of the ion channel in the cell membrane. It describes how action potential initiated and propagated through neurons, the neuronal unit of communication. Neuronal Modelling can be computationally expensive, specially when modelling with the high resolution level models. It becomes even more challenging when considering tuning many parameters that changes with the biological properties of each sample, making the large-scale modelling big challenge in the field. Neural differential equations can propose a promising direction as data-driven differential solvers, these models can combine the current advance of machine learning with the domain knowledge of the systems. In this project, neural differential equation models are represented to solve Hodgkin-Huxley equations by combination of neural networks Approximators for gating variables of ion channels and the differential equation of how voltage is changing cross cell membrane.

This project implemented the oscillatory network as presented by Matsuoka et al in order to control a bipad walking robot and snake crawling robot in two different challenging tasks. Central pattern generators (CPGs) are self-organizing biological neural circuits that produce rhythmic outputs. These output can be used to control of locomotion in bio-inspired robots. This mechansim can bring interesting insights on how to gain inspiration from the biological systems in order to build more robust and natural behavouir in robots’ movement. The project implemented a CPG model in microcontroller chip using arduino and connceted it to control the robots to perform the optimised outputs for two tasks.

In this project, we built a system to transmit data between rooms via local WLAN networks using an IMU for encoding and four vibrotactile motors for decoding. we proposed a new method combining a threshold-based encoding strategy using IMU and EMG capable of encoding the numbers 0-9 to maximize the information transfer rate (ITR) without heavy pre-training for participants. An EMG signal is used as error-correction source to improve the accuracy on the encoding side. In experimental trials maximal information transfer rates of above 100 bit per minute and more than 99% accuracy were achieved in trials optimizing for one of both performance measurements at a time.

Hippocampal sharp wave ripples have been identified as key biomarkers of important brain functions such as memory consolidation and recall of episodic memory. In order to understand the causal relationship between the mechanism of the sharp waves and behaviour. The SWR events can be disturbed by electrical stimulation. This is rely on accurate real-time SWR detection system. This project is aimed at building a real-time system for detecting sharp wave ripples that can be used in a closed-loop perturbation experiments, i.e future optogenetics researches. The designed model can achieve less than 20ms latency, which less than the speed duration of generating and vanishing of the sharp-wave ripples signal, so it can detect it and disturb it before it disappears. With around 97% accuracy makes it a reliable model to be implemented in the closed-loop system.

The need to find scalable, rapid and accessible, yet accurate, methods of diagnosis is crucial for fighting pandemics. The COVID-19 infected patients may represent on chest X-ray images with a pattern that is only barely characteristic for the human eye, Accordingly; machine learning can play a critical role in automating patterns recognition faster, and more precise. The model trained on a pre-classified database of X-ray images from COVID-19, Pneumonia, and healthy patients. The model is designed using Convolutional Neural networks, with ResNet architecture, and transfer learning by pre-trained the model on ImageNet. It is been trained to recognize the patterns of the three classes (COVID-19, non-COVID-19 Pneumonia, or Normal) from 3429 X-ray samples, and its out-of-distribution performance examined using 378 X-ray images of different patient, succeeded in classifying the predicted class correctly by accumulated accuracy of (~92%).

Aimed at improving the quality of life for amputees after the devastating war in my city, this project designed and built an EMG-based prosthetic hand that operates as a part of the human body’s nervous system, to participate in healing the current tragedy in our society. Our design was implemented by acquiring EMG data using local-made electrode and converted it into mechanical motion according to the desired 6 different movements. This model manipulates the electrical pulses by extracting some features -using Statistical features and Wavelet techniques- classified using machine learning into 6 multi-classes depending on their functionalities. The model is operated by servo motors which are controlled by Arduino Uno.