Abstract

The lung infection caused by COVID-19 constitutes a medical emergency, necessitating prompt detection and treatment to mitigate its effects. Clinical diagnosis and assessment of severity are typically conducted utilizing medical imaging practices, like chest X-rays and lung CT scans. Owing to its advantages, Deep-Learning (DL) frameworks are extensively developed to identify occurrences utilizing clinically obtained X-ray/CT data. This study plans to propose a novel framework for COVID-19 detection in X-ray/CT images utilizing the Lightweight Deep-Learning Model (LDLM). This framework included the LDLM features alongside the Local Binary Pattern (LBP) features to enhance detection efficacy. Additionally, to reduce the over-fitting problem, this framework utilized the Arithmetic-Optimization Algorithm (AA) for features reduction and fusion to enhance detection accuracy. The stages of this framework encompass; data collecting and preliminary adjustment, feature extraction utilizing LDLM, and LBP, feature selection by AA and serial feature concatenation, and classification employing five-fold cross-validation. The suggested framework was evaluated and validated using the benchmark image database, yielding accuracy >99% for both X-ray and CT based examinations. These results validate that the proposed approach yields substantial outcomes on the selected benchmark database. 

Statistics