Hybrid Metaheuristic Optimisation for Lung Cancer Image Classification: Leveraging MOEA, PSO, and ACO Algorithms

In the critical domain of cancer classification, accurately identifying lung cancer scans as malignant or benign is paramount. This study proposes a novel approach that leverages a hybrid metaheuristic combination of algorithms. Ant Colony Optimization (ACO), Particle Swarm Optimization (PSO), and the Knapsack Algorithm are utilized for feature selection. Subsequently, Convolutional Neural Networks (CNNs), recognised as state-of-the-art in medical image classification, are employed. Among the three algorithms, ACO performed the best. Following this, Multi-Objective Evolutionary Algorithms (MOEAs), Particle Swarm Optimisation, and the Simplex Algorithm were used, respectively, to fine-tune hyperparameters and enhance Classification accuracy. The best result was obtained by using ACO for feature selection and PSO for hyperparameter tuning, achieving an accuracy of 98.83%. The study conducts a comparative analysis of different algorithmic combinations to validate the proposed methodology’s efficacy. This comparative analysis of optimisation algorithms significantly maximises the accuracy of lung cancer classification.