A Modified Chicken Swarm Optimisation Algorithm for Feature Selection of a Multi-Textural Feature Digital Image Classification System

Digital image processing is a process that involves analyzing and manipulating images digitally via computer which has various applications such as remote sensing, surveillance, Biometrics, Medical field and more. Brain tumours are diseases that occur in the brain when abnormal cells begin to develop in an uncontrolled manner. The growth could be fatal and deadly if the accumulation continues. Thus, the quick discovery of the brain tumour is significant and helpful for further investigation. Classification and identification are challenging due to image complexity and unclear causes. This paper proposes a modified Chicken Swarm Optimisation (mCSO) technique for feature selection in digital images classification. 1800 brain MRI images was acquired from the Kaggle database. The brain tumour dataset were preprocessed. Three techniques (gray-level co-occurrence matrix, discrete wavelet transformation, and Gabor filter) were used for feature extraction and their outputs were fused by Serial Sum technique. The Chicken Swarm Optimisation was modified by Simulated Binary Crossover to prevent its local optima problem. The result of the analysis is focused on multi-binary classification to determine the efficacy of fusing feature extraction methods. The study found that the technique with mCSO achieved an accuracy of 97.61% better than the standard Chicken Swarm Optimisation technique that achieved accuracy of 96.50%.