A Novel Deep Learning-Based Classification Framework for COVID-19 Assisted with Weighted Average Ensemble Modeling.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Infectious Disease Diagnosis

Background:

• COVID-19, caused by SARS-CoV-2, presents diverse symptoms and requires timely diagnosis to prevent severe lung complications.
• Current diagnostic methods like RT-PCR are accurate but time-consuming and labor-intensive.
• Deep learning on medical imaging offers automated COVID-19 detection, but existing systems face limitations like overfitting and generalization errors.

Purpose of the Study:

• To develop a highly accurate, efficient, and reliable deep learning-based technique for COVID-19 detection.
• To address limitations of existing automated systems by employing transfer learning and improved preprocessing.
• To enhance the reliability of COVID-19 detection through an ensemble approach.

Main Methods:

• An ensemble deep learning model was created by combining Xception, VGG19, and ResNet50V2 Convolutional Neural Network (CNN) models.
• A weighted average ensemble (WAE) prediction strategy was utilized.
• Transfer learning and enhanced preprocessing techniques were applied to two benchmark datasets.

Main Results:

• The WAE model achieved 97.25% accuracy for binary classification and 94.10% for multiclass classification of COVID-19.
• The proposed method demonstrated higher accuracy compared to single CNN models.
• The use of transfer learning and advanced preprocessing improved the reliability of the detection system.

Conclusions:

• The proposed ensemble deep learning technique provides a reliable and accurate method for automated COVID-19 detection.
• This approach overcomes limitations of traditional methods and existing deep learning models.
• The study highlights the potential of ensemble learning and transfer learning in medical image analysis for disease diagnosis.