A Glimpse of 3D/2D Chest CT Feature Fusion

Prompt diagnosis and severity assessment of respiratory infections in 3D chest volumes of computed tomography (CT) scans is crucial for improving their prognosis. This study introduces an innovative Integrative Computer-Aided Framework (ICAF) to gauge the severity of lung infections from 3D chest scans. ICAF integrates volumetric 3D and traditional 2D features. These 3D features are derived using a designed 3D-Convolution Neural Network (3D-Norm-VGG16) model inspired by literature, while 2D features are obtained via a fuzzy-based Region of Interest (RoI) segmentation. ICAF effectively identifies ground glass opacities (GGO) within 3D CT volumes, offering improved severity analysis. ICAF is therefore applied to categorize the severity of COVID-19. Comprehensive testing on datasets such as MosMedData delivers outstanding performance. It achieved an overall accuracy of 98.40 %, average precision, recall, and F1-score of 98.04 %, 97.87 %, and 97.95 %, respectively, surpassing state-of-the-art models.