ET-NET

ET-NET applies ensemble transfer learning to classify COVID-19 from lung CT-scan images.

Key Features:

• Automated screening: Deep learning analysis of chest CT-scan images to detect COVID-19-related patterns.
• Ensemble learning (Bagging): Bootstrap aggregating combines predictions from multiple models to reduce variance and improve reliability.
• Transfer learning with pre-trained networks: Integrates Inception v3, ResNet34, and DenseNet201 as constituent models in the ensemble.
• Performance metrics: Reported metrics on a public dataset using 5-fold cross-validation: accuracy 97.81% ± 0.53%, precision 97.77% ± 0.58%, sensitivity 97.81% ± 0.52%, specificity 97.77% ± 0.57%, outperforming prior methods on the same dataset by 1.56%.

Scientific Applications:

• COVID-19 screening from CT scans: Automated classification of lung CT-scan images to identify cases consistent with COVID-19 infection.
• Clinical diagnostic support: Supplementing clinical decision-making where CT imaging is used to assess suspected COVID-19 cases and prioritize resources.

Methodology:

Uses transfer learning with pre-trained Inception v3, ResNet34, and DenseNet201 models combined via bootstrap aggregating (bagging) ensemble and evaluated with 5-fold cross-validation on a publicly available CT dataset.