A novel LT-LBP based prediction model for COVID-CT images with Machine Learning

Logarithmic Transformation of Local Binary Pattern (LT-LBP) is introduced with machine learning algorithms for the classification of covid CT images. Preprocessing the input information plays a key role in machine learning as well as deep learning models. CT images and Chest X-rays are significant in diagnosing the disease. Most of the medical images are greyscale images. Texture analysis is one of the ways to obtain information from medical images and Local binary pattern is an efficient texture operator. With the texture pattern LBP, a novel method as Logarithmic Transformation of Local Binary Pattern (LT-LBP) is proposed in this paper. We applied 2695 CT images and 115 Italian COVID Positive CT images and 5 COVID positive CT images from the Chennai region. The CT-Scans and Chest X-ray images have endured for preprocessing and texture analysis with Local Binary Pattern (LBP) and trained with Support Vector Machine (SVM), K-nearest neighbors (KNN), Random Forest (RF), Logistic Regression (LR) as machine learning algorithms. The LT-LBP gives a better result when compared with normal LBP when combined with SVM and RF. The retrospective study gives the result as accuracy percentage of 95.7 with LT-LBP combined with SVM and also 91.4 percent of accuracy results for LT-LBP with RF