A FUSION OF IMAGE PROCESSING AND DEEP LEARNING FOR COVID19 DETECTION USING 2D ITERATIVE CONVOLUTIONAL NEURAL NETWORK

Abstract

Covid-19 still continues to be cataclysmic danger to humankind even after the discovery of vaccines because of passing of similar mutants which leads to creation of new variants. Image processing techniques are fused with a deep learning model to bring out the detection of covid19. A Raw Low Dose CT database Images (RLD-CTDI) are used along with the CAD approach to bring out a novel automatic framework. Raw Ct images in general have some clamors such as Gaussian, pepper & salt; speckle noises etc or might even be affected by shaky voltage disturbance. To remove these clamors and disturbances 2D Improved Anisotropic Diffusion Bilateral Filter (2D IADBF) is used which restores the image. The image is further pre-processed by using 2D Edge Preservation Efficient Histogram Processing to preserve the edges. After the pre-processing steps a clear noise-free image is obtained for further processing like clustering and thresholding. Clustering is done using 2D Hybrid-Fuzzy C-Means Algorithm (2D HFCM) to obtain disease clusters and thresholding is done using 2D Adaptive OTSU Thresholding to extract the Region of Interest (ROI). Using the ROI, Feature extraction is applied using Gray-Level Co-Occurrence Matrix Histogram Of Gradient (GLCM HOG) calculation to obtain features. These features are fed as input to the deep learning model.2D Iterative Convolutional Neural Network is used for classification of the image which categorizes the CT image into covid affected / Non-covid affected image.