Related Experiment Video
Updated: Jul 24, 2025
04:17
DNA Virus Detection System Based on RPA-CRISPR/Cas12a-SPM and Deep Learning
Published on: May 10, 2024
818
Automated COVID-19 detection with convolutional neural networks.
Aphelele Dumakude1, Absalom E Ezugwu2
1School of Mathematics, Statistics, and Computer Science, University of KwaZulu-Natal, King Edward Avenue, Pietermaritzburg Campus, Pietermaritzburg, 3201, KwaZulu-Natal, South Africa.
Scientific Reports
|June 30, 2023
Summary
This study introduces two AI models for automated COVID-19 detection using chest CT scans. Both models achieved high accuracy, outperforming existing methods for efficient and reliable screening.
Area of Science:
- Artificial Intelligence in Medical Imaging
- Machine Learning for Disease Detection
- Computational Pathology
Background:
- The critical need for rapid and precise automated screening tools for COVID-19 detection.
- Existing research in automated diagnostic frameworks for infectious diseases.
- Challenges in achieving high accuracy and efficiency in COVID-19 screening using medical imaging.
Purpose of the Study:
- To develop and evaluate two novel framework models for automated COVID-19 detection from chest CT scans.
- To compare the performance of a hybrid CNN-XGBoost model against a CNN-Feedforward Neural Network model.
- To establish a benchmark against state-of-the-art methods using the CovidxCT-2A dataset.
Main Methods:
- Development of two distinct deep learning architectures: a hybrid CNN-XGBoost model and a CNN-Feedforward Neural Network model.
- Application of Bayesian optimization for hyperparameter tuning and transfer learning techniques (Dropout, Batch Normalization) to prevent overfitting.
- Training, validation, and testing conducted on the CovidxCT-2A dataset, with performance evaluated using Precision, Recall, Specificity, Accuracy, and F1-score.
Main Results:
- The hybrid model achieved superior performance with 98.43% precision, 98.41% recall, 99.26% specificity, 99.04% accuracy, and 98.42% F1-score.
- The standalone CNN model also demonstrated strong results: 98.25% precision, 98.44% recall, 99.27% specificity, 98.97% accuracy, and 98.34% F1-score.
- Both proposed models significantly outperformed five other state-of-the-art models in classification accuracy.
Conclusions:
- The developed AI frameworks offer efficient and accurate automated screening for COVID-19 detection using chest CT images.
- The hybrid CNN-XGBoost model shows particular promise for clinical application due to its high performance metrics.
- These findings contribute to advancing AI-driven diagnostic tools for public health emergencies.
