Advanced Detection of COVID-۱۹ Delta Variant in Chest CT Scans Using a ConvXGBoost Model with Deep Transfer Learning

The emergence of new coronavirus variants, such as Delta, continues to cause global crises, highlighting the need for rapid diagnostic methods to control the virus's spread. Chest Computed Tomography (CT) has proven to be a reliable detection and assessment tool for COVID-۱۹. This study aimed to develop an accurate artificial intelligence method to assist clinicians in identifying COVID-۱۹ patients through CT images. A dataset of ۱۰۳ chest CT scans was collected from an imaging clinic in northern Iran, comprising ۷۵ scans from Delta variant-infected patients and ۲۸ from healthy individuals for comparison. The methodology included two image reconstruction steps: a fuzzy enhancement technique and stacking the enhanced results with the original images to aid model validation. A novel model, ResNet۵۰V۲XGBOOST, was developed by combining Convolutional Neural Networks (CNN) with the XGBoost classifier. Using transfer learning, several pre-trained models, including InceptionV۳, Xception, MobileNetV۲, and ResNet۵۰V۲, were evaluated to select the optimal one. The experimental results demonstrated that the ResNet۵۰V۲XGBOOST model achieved an accuracy of ۹۹.۵۰%, sensitivity of ۹۸.۹۸%, and specificity of ۹۹.۶۳% in identifying COVID-۱۹ patients.

کلیدواژه ها:

COVID-۱۹ Detection ، Deep Learning ، Transfer Learning.