Breast cancer diagnosis using hybrid AlexNet-ELM and chimp optimization algorithm evolved by Nelder-mead simplex approach

This study proposes a Hybrid AlexNet-Extreme Learning Machine (ELM) approach for breast cancer diagnosis using mammography images. Batch normalization is applied to improve AlexNet's performance, and the chimp optimization algorithm (ChOA) is utilized to avoid sub-optimal solutions in ELM. The Nelder-mead simplex (NEMS) technique is then employed to enhance the convergence behavior of ChOA. The study's main contributions are the proposed hybrid model and the application of ChOA and NEMS techniques to improve the performance of ELM. The proposed model is evaluated using the CBIS-DDSM dataset, with wiener and CALHE filters used as preprocessors. The effectiveness of the classification is examined using five optimization algorithms, and several metrics. The outcomes demonstrate that CALHE filter offered the best performance overall, and AlexNet-BN-ELM-CHOA-NEMS was the most accurate of the five models, with a sensitivity of 96.03 %, a specificity of 94.60 %, and an overall accuracy of 95.32 %. The findings demonstrate the effectiveness of the proposed model in breast cancer diagnosis.