Hybrid Marine Predators and Parrot Optimization Approach For Mammographic Mass Classification

Abstract

Breast cancer is regarded as among the most serious health issues impacting women globally. Mammography is the benchmark standard for diagnosing this deadly disease, but the inherent subjectivity of human interpretation may introduce biases that adversely affect outcomes. In recent years, deep learning-based medical image analysis techniques have emerged to assist clinicians in precise disease diagnosis. In this research, a Chaotic Binary Marine Predators Algorithm integrated with Parrot Optimization (MPPA) model, which uses chaotic maps in order to provide better exploitation and exploration balance throughout the search process, is proposed. To conduct the research, regions of interest (ROIs) are extracted from the CBIS-DDSM mammogram image dataset and are pre-processed using various steps. Further, the deep features are extracted using the AlexNet and ResNet model and optimised using the proposed Hybrid versions. The proposed algorithm aims to enhance classification accuracy while reducing feature dimensionality. In this regard, an extreme learning machine (ELM)- based fitness function is considered. Further, the performance of the proposed algorithm is also validated on standard datasets from the UCI repository. Experimental results signify that the proposed algorithm outperforms the other algorithms considered for performance comparison for the efficient classification of breast mass in mammogram images.