3D Convolutional Neural Network and K-Nearest Neighbour with Radiomics Features for Covid Classification

Abstract

For the purpose of identifying COVID-19 lung affection When it comes to finding tumors, computed tomography plays a crucial role. However, the intricacy of the CT makes categorization a formidable challenge for the doctor. From 2D CT, just a handful of characteristics may be retrieved. Compared to 2D CT, 3D CT offers additional features and equivalent diagnostic performance. Learn how to spot and categorize a COVID-19 lung ailment with the use of 3D computed tomography (CT) in this article. To acquire a plethora of details from CT scans, radiologists utilize data-characterization algorithms. Discovering illness traits is possible with the use of these radiomics properties. To extract GLCM features, we utilize pyradiomics, an open-source library for Python. The 3D CT data is processed using a combinations model that incorporates GLCM, CNN, and KNN features. To get a stronger three-dimensional volumetric representation, 3D-CNN is used. It is the job of the final 3D-CNN layer to learn an accurate image representation. From that layer, we extract the characteristics and feed them into the KNN classifier so it can make more predictions. This approach has been shown to enhance accuracy by as much as 98.9%.