A HYBRID CNN AND XGBOOST APPROACH FOR ACCURATE COVID-19 PNEUMONIA DETECTION USING CHEST X-RAYS

Abstract

The worldwide spread of COVID-19 has made things very hard for doctors, especially when it comes to quickly and accurately diagnosing COVID-19 pneumonia from chest X-rays. Traditional ways of diagnosing are often time-consuming and need a lot of medical knowledge, which can cause delays in treating patients. The main issue this study tries to solve is the need for a fast and accurate automatic method that can make finding COVID-19 pneumonia easier and more reliable. To make diagnostics work better, this study suggests a mixed classification system that combines Convolutional Neural Networks (CNN) with the XGBoost method. The main goal of this study is to create and test a strong model that does a better job of classifying COVID-19 pneumonia than existing machine learning and deep learning methods. The standard dataset used for research work as "Chest X-ray COVID-19 Pneumonia." It has 6432 chest X-ray images spread out across three classes. The suggested way combines CNN models for feature extraction, such as EfficientNetB0 and DenseNet121, with fine-tuned XGBoost algorithms to help make better decisions. The suggested hybrid model was shown to be more accurate and stable through extensive testing and comparisons with other models, such as SVM and Random Forest. There is a lot of evidence that this method works really well for helping doctors make decisions about COVID-19 tests.