Evolutionary Tool for Denoising DNA Microarray Images Using CNN

Abstract

Deoxyribonucleic acid (DNA) microarray technology has promised rapid improvement in recent studies. On DNA microarray images, there are several spots. Spots on microarray images represent gene expression and show the status of normal and malignant cells. One of the approaches for enhancing and analysing an image is to use digital image processing. In this paper, a convolutional neural network (CNN)-based denoising technique is developed and deployed. It is based on feature extraction from images. It is no longer necessary to manually extract features. The simulation is carried out using MATLAB to evaluate the performance parameters like peak signal-to-noise ratio (PSNR), mean squared error (MSE) and structural similarity index (SSIM). The comparison between different filtering techniques and CNN is carried out checking for the efficiency. The traditional image noise reduction methods such as median filter method, linear filter method and Weiner filter methods are compared with CNN. The advantage is that the CNN model’s parameters may be fine-tuned by network testing and training, whereas the parameters of traditional image denoising methods are fixed and cannot be changed throughout the filtering process and resulting in a lack of configurability.