An Advanced Ensemble of Deep Learning Models for Breast Cancer Segmentation and Classification with Two-Tier Optimization Algorithms
M. Sreevani, R. Latha · Engineering Technology & Applied Science Research · 2025
Breast Cancer (BC) is one of the most common cancers among women. Routine mammography is substantial because asymptomatic BC does not show early signs, making early detection difficult. Automated methods, including Deep Learning (DL) models, have gained significant attention for analyzing mammographic images and enhancing diagnostic accuracy. Successful AI training for these medical tasks depends on large datasets with accurately annotated lesion locations. This study proposes an Advanced Ensemble Deep Learning Model for Breast Cancer Segmentation and Classification with a Two-Tier Optimization (AEDL-BCSCT2O) approach to segment and classify BC using advanced DL and optimization techniques. The model initially applies Adaptive Bilateral Filtering (ABF) for noise removal and CLAHE for contrast enhancement to improve image quality. The DeepLabV3+ segmentation method is enhanced through parameter optimization using the Lemur Optimizer (LO). The NASNetMobile model is utilized for feature extraction. An ensemble of Deep Belief Network (DBN), Graph Convolutional Network (GCN), and Sparse Stacked Autoencoder (SSAE) models is used for improved classification. Finally, the Osprey Optimization Algorithm (OOA) approach is utilized for tuning. The validation results show that the AEDL-BCSCT2O method achieves 99.76% accuracy, outperforming existing models.