Enhancing Diagnostic Accuracy with Ensemble Techniques: Detecting COVID-19 and Pneumonia on Chest X-Ray Images

Abstract

Lung diseases such as COVID-19 and pneumonia can lead to breathing difficulties, decreased lung function, and respiratory failure, leading to death if not treated immediately. Chest X-ray imaging techniques are quick, effective, and inexpensive in controlling many of these diseases. Artificial intelligence has shown promising results in detecting many diseases, including lung diseases, as it can analyze large sets of data in a short time, which leads to causes reducing the spread of these diseases. Artificial intelligence with the help of deep learning and machine learning is a milestone in modern biomedical research, offering major advances in various medical research disciplines. In this research paper, we introduced various advanced ensemble techniques as bagging, boosting, stacking, and blending with different algorithms, to enhance the performance of our classification models in detecting coronavirus and pneumonia. We specifically focused on combining convolutional neural network (CNN) and vision transformer (ViT) models to create powerful ensemble models. Our objective was to determine the most accurate ensemble technique for diagnosing lung diseases. We assessed their ability to correctly classify chest X-ray images as either COVID-19, pneumonia, or normal. The CatBoost algorithm achieved the highest accuracy of 99.753% using the COVID-19 Radiography dataset and the bagging ensemble model achieved the highest accuracy of 95.08% using 5 different publicly available data (COVIDx CXR-4). The results indicate that the advanced ensemble techniques can significantly improve the performance of machine learning models. [JJCIT 2024; 10(4.000): 428-442]

Keywords

• covid-19
• cxr
• densenet-169
• vit-l32
• stacking
• boosting
• bagging
• blending