Transformative Insights: Gabor Features and Tensor-EWT for COVID Diagnosis in Lung CT Images

The 2019 outbreak of the novel coronavirus (COVID-19) significantly affected the advancement of the global economy and the quality of life for people This paper addresses the limitations of traditional and trending tools in accurately and promptly detecting COVID through CT images, considering the highly contagious and dangerous nature of the illness caused by the SARS-CoV-2 virus. Therefore, the paper focuses on developing a solution that automatically detects COVID. The novel approach is based on the Tensor-Empirical Wavelet Transform (TEWT), which is proposed to decompose diagnosed pre-processed images. Significant Gabor-based features are extracted from decomposed sub-band images, and relevant features are recognized using Neighborhood Component Analysis (NCA). Then, robust features are ranked using the Student's t-value algorithm, and ranked features are fed into LS-SVM for classification. The model achieved an impressive 96.29% increase in classification accuracy, improvements of 96.56% in specificity, 96.03% in sensitivity, and an F1 score of 95% using 5-fold cross-validation. The study's outcome is to demonstrate that the model exhibits state-of-the-art COVID classification techniques.