Use of Novel Artificial Intelligence for the Development of Mathematical Model
Loading...
Date
item.page.authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
newline Abstract
newlineThe ongoing COVID-19 pandemic has underscored the urgent need for rapid and accurate
newlinediagnostic tools. Among these, the precise classification of chest X-ray images into COVID
newline19 and non-COVID-19 cases plays a pivotal role in effective disease management and
newlinecontrol. Traditional machine learning algorithms and Convolutional Neural Networks
newline(CNNs) have been employed for this task, but they often face challenges related to accuracy,
newlineprecision, and computational efficiency. Striking a balance between these crucial metrics
newlineremains an ongoing challenge in the field of medical image analysis.
newlineThis thesis explores the field of optimizing Graph Neural Networks (GNNs) through a
newlinerigorous mathematical analysis. It introduces a novel framework designed to address the
newlinelimitations of existing methods. This framework leverages various mathematical techniques,
newlineincluding spectral methods, dynamic graph sparsification, game-theoretic attention
newlinemechanisms, Bayesian uncertainty models, and advanced graph partitioning methods. By
newlineincorporating these elements into the GNN architecture, this research aims to significantly
newlineenhance the classification of COVID-19 chest X-ray images.
newlineThe need for such optimization arises from the shortcomings of current approaches. While
newlinetraditional machine learning algorithms and CNNs have been effective to some extent, they
newlineexhibit limitations in terms of sensitivity, specificity, and computational speed. These
newlineconstraints impede their practical utility in healthcare settings where timely and accurate
newlinediagnoses are paramount.
newlineIn response to these challenges, this thesis offers a multifaceted approach to improve
newlineclassification metrics across various dimensions. The incorporation of spectral methods
newlineallows for a more comprehensive understanding of graph structures, enabling GNNs to
newlinebetter capture essential features in the data. Dynamic graph sparsification techniques help
newlinestreamline the computational process, reducing unnecessary complexity while maintaining
newlinec