Towards Effective COVID-19 Medical Image Classification Using Metaheuristics and Machine Learning Approaches
Diagnosing life-threatening diseases like COVID-19 requires accurate and efficient methods to support timely decision-making. This research explores this possibility by introducing a new framework that combines optimisation algorithms with deep learning techniques. Hybrid optimisation methods were developed by pairing Wavelet Entropy with Artificial Bee Colony, Genetic Algorithm, and Particle Swarm Optimisation. Among these, WE-PSO achieved the best performance. It delivered an accuracy of 90.97% and an F1 score of 85.67%, effectively handling complex medical datasets. To address the limitations of traditional optimisation methods, a new model was created. This model combines the lightweight architecture of SqueezeNet with the classification capabilities of a Gaussian-kernel Support Vector Machine. It is referred to as the SGS model. The design balances efficiency and performance, making it suitable for resource-constrained environments. Experimental results demonstrated the superiority of the SGS model. It achieved an accuracy of 98.49% and an F1 score of 98.18%. This significantly outperformed WE-PSO and other methods. This research establishes a robust framework for COVID-19 detection and opens new possibilities for advancements in medical diagnostics.
History
Supervisor(s)
Fuxiang Chen; Shiqiang YueDate of award
2024-12-12Author affiliation
School of Computing and Mathematical SciencesAwarding institution
University of LeicesterQualification level
- Masters
Qualification name
- Mphil