Performance Analysis Of Artificial Intelligence Based Asymptomatic Breast And Lung Cancer Classification Using Microarray Gene Expression Data

Abstract

ABSTRACT newlineSince the previous decade, the frequency and fatality of several forms of cancer newlinehave increased. Because of the unrestricted growth and metastasis of cancer newlinecells, conventional image processing-based CAD systems are ineffective in newlineclassifying cancer types. Furthermore, present CAD systems identify cancer at newlinean advanced stage, which causes issues in cancer therapy since it relies on newlinespecific symptoms such as a tumor in the medical image. Another disadvantage newlineof the current image-based cancer classification CAD system is its focus on a newlineparticular cancer type. The same medical image cannot be used to represent newlinevarious cancer conditions. Using Deoxyribo Nucleic Acid (DNA) microarray newlinetechnology, gene expression profiles allow for categorizing cancer types at an newlineearly stage, i.e., at the molecular level. It aids in the effective detection of newlineasymptomatic and various cancer conditions. However, performing microarray newlinegene expression data processing is difficult due to its complexity, profile newlineredundancy, gene feature redundancy, and inadequate gene processing methods. newlineIn this thesis, novel microarray gene expression data analysis methods were newlineproposed using advanced Artificial Intelligence (AI) algorithms for the newlineclassification of asymptomatic breast and lung cancers. This thesis proposed newlinefour contributions of efficient gene profile analysis for the Asymptomatic newlineDisease Classification (ADC) system. newlineIn the first contribution, the dynamic gene data pre-processing algorithm was newlineproposed to dynamically identify the redundant gene profiles and discard them. newlineAdditionally, the low-expressed gene profiles were removed which helps the newlineADC performance improvement. Experimental results using breast and lung newlinegene datasets proved that the proposed pre-processing technique outperformed newlineexisting methods. In the second contribution, the dynamic data pre-processing newlinetechnique is applied to different gene Feature Selection (FS) methods for ADC. newlineThe gene FS methods such as Fisher Score (FS), Relief F (R