Related Experiment Video
Updated: Aug 16, 2025
14:27
Identification of Disease-related Spatial Covariance Patterns using Neuroimaging Data
Published on: June 26, 2013
15.7K
PSTCNN: Explainable COVID-19 diagnosis using PSO-guided self-tuning CNN
Wei Wang1, Yanrong Pei2, Shui-Hua Wang1
1School of Computing and Mathematical, University of Leicester, Leicester, LE1 7RH, UK.
Summary
This study introduces a novel Particle Swarm Optimisation-guided Self-Tuning Convolution Neural Network (PSTCNN) for faster and more effective COVID-19 diagnosis. The PSTCNN model automatically tunes hyperparameters, reducing human involvement and improving diagnostic accuracy.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Computational Biology
Background:
- Coronavirus disease-19 (COVID-19), caused by SARS-CoV-2, presents a significant global health and economic challenge.
- Accurate and rapid diagnosis of COVID-19 is crucial but strained by limited healthcare resources.
- Deep learning models offer potential for computer-aided diagnosis but require precise hyperparameter tuning for optimal performance.
Purpose of the Study:
- To develop an automated hyperparameter tuning method for deep learning models in COVID-19 diagnosis.
- To introduce the Particle Swarm Optimisation-guided Self-Tuning Convolution Neural Network (PSTCNN) for efficient and accurate COVID-19 detection.
- To reduce human intervention and improve the stability and speed of the diagnostic model training process.
Main Methods:
- Implementation of Particle Swarm Optimisation (PSO) to guide the self-tuning process of a Convolution Neural Network (CNN).
- Development of the PSO-guided Self-Tuning Convolution Neural Network (PSTCNN) for automated hyperparameter optimization.
- Experimental validation of the PSTCNN model on COVID-19 diagnosis datasets.
Main Results:
- The PSTCNN achieved high performance metrics, including sensitivity (93.65%±1.86%), specificity (94.32%±2.07%), accuracy (93.99%±1.78%), and F1-score (93.97%±1.78%).
- The optimization algorithm enabled targeted selection of hyperparameters, leading to stable solutions near the global optimum.
- Compared to traditional methods, the PSTCNN demonstrated faster and more effective hyperparameter tuning.
Conclusions:
- The PSTCNN model offers an effective and automated solution for hyperparameter tuning in deep learning-based COVID-19 diagnosis.
- Automated tuning using optimization algorithms like PSO significantly enhances diagnostic model efficiency and accuracy.
- The proposed approach alleviates pressure on healthcare systems by providing a reliable computer-aided diagnostic tool.
