Authors:
M. A. Sayedelahl, R. M. Farouk
Addresses:
Department of Computer Science, Damanhour University, Damanhour, Egypt, mohamed.abdelatif@cis.dmu.edu.eg. Department of Mathematics, Zagazig University, Zagazig, Egypt, rmfarouk1@yahoo.com.
Accurate and precise segmentation of microarray images is crucial for reliable quantification of gene expression levels, a cornerstone of modern genomic research. We propose a novel supervised classification approach to effectively discriminate between foreground and background regions within microarray images to address this critical challenge. Our method incorporates Gabor filter-based preprocessing to enhance image features and automated spot localization to identify regions of interest. Subsequently, a comprehensive set of features is extracted from each pixel to capture relevant image characteristics. A Radial Basis Function (RBF) network is then employed to classify pixels as either foreground or background. Rigorous evaluation of real microarray datasets from SMD and UNC demonstrates the superior performance of our method compared to conventional techniques such as k-means clustering and Support Vector Machines, achieving a notable improvement of up to 20% in segmentation accuracy. The enhanced segmentation accuracy provided by our approach has the potential to significantly improve the reliability and precision of downstream gene expression analysis, contributing to advancements in genomic research and discovery.
Keywords: Microarray Image Analysis; Microarray Image Segmentation; Gabor Filter; Spot Segmentation; Radial Basis Function Network; Support Vector Machines (SVMs); Recurrent Bayesian Neural Network (RBNN).
Received on: 10/11/2023, Revised on: 02/01/2024, Accepted on: 05/03/2024, Published on: 07/06/2024
AVE Trends in Intelligent Computing Systems, 2024 Vol. 1 No. 2, Pages: 77-90