Artificial Neuron Model for Large Scale Spiking Brain Neural Network

Asif Mehmood

PhD thesis, University of Engineering and Technology, Taxila, Pakistan, 2025

Doctoral thesis, Department of Computer Science, Faculty of Telecommunication and Information Engineering, University of Engineering and Technology, Taxila, Pakistan. Supervisor: Dr Muhammad Javed Iqbal. The PDF is the book edition of the thesis (132 pages, including the MATLAB source code as Appendix A).

Abstract

Modeling and simulation of human brain with diverse neurocomputational features is a challenging task. To perform biologically meaningful brain simulation, biologically comprehensive and computationally simple model of neuron is required. State of the art models in literature are either complex in terms of biological details or exhibit limited neurocomputational features. The essential part of information processing in brain is concerned with diverse neurocomputational features and bidirectional communication between brain and organs. In this work a unified model of excitable cells is proposed. The model contains the dynamics of neurotransmitters to replicate the enriched neurocomputational features and adrenergic receptors to handle the bidirectional communication between brain and organs. A network of 2191 fully connected neurons, as per the dimensions of dataset, is simulated and performance of proposed model is compared with Generalized Leaky Integrate and Fire (GLIF) model, Multi-scale Adaptive Threshold (MAT) model, and Quadratic Integrate and Fire (QIF) model. It is observed that proposed model is 86% efficient in terms of coincidence factor when compared with GLIF 69%, MAT 73%, and QIF 77%. In terms of firing rate coefficient, prediction accuracy of proposed model is 83% against GLIF 55%, MAT 61%, and QIF 67%. The model extended to the neural network of human brain level containing 100 billion neurons by assuming the neural response in digital form. This assumption reduced the memory requirement of simulation to 13 gigabytes. The model applied to the cells of heart and successfully replicated the electrical activity of cardiac myocyte of different species. A network of 1000 fully connected neurons and cardiac myocytes simulated, and electrical activity is recorded. From electrophysiological data of cardio-neural network simulation, heart rate variability is predicted. This study provides the modeling and simulation phase of brain-heart interface to predict morbidity at early stages. The recent advancements in nano-electronics will make is possible to develop brain-heart interface as nano-chip implant in subject to monitor and control the brain-heart interplay through electrophysiological signals.

Keywords

• Spiking neural networks
• Hybrid excitable cell model
• Action potential
• Adaptive threshold
• Cardiac myocyte
• Brain–heart interface

Cite this work

@phdthesis{mehmood2025thesis,
  title  = {Artificial Neuron Model for Large Scale Spiking Brain Neural Network},
  author = {Mehmood, Asif},
  school = {University of Engineering and Technology, Taxila},
  address= {Taxila, Pakistan},
  year   = {2025}
}