Breast Cancer Segmentation in Mammograms using Antlion Optimization and CNN/GRU Architectures

Early and accurate diagnosis of breast cancer is crucial for successful treatment and improved patient outcomes. This paper proposes a novel hybrid approach for breast cancer classification in mammographic images that combines the powerful optimization capabilities of Antlion Optimization (ALO) with the feature extraction and learning potential of a deep learning (DL) model. To enhance breast cancer detection, we introduce a novel preprocessing approach that combines a Gaussian filter with Residual Pixel Removal (RPR). This initial step aims to reduce noise and highlight relevant features in the images. We then extract textural features using the Gray Level Cooccurrence Matrix (GLCM), providing valuable insights into the spatial distribution of intensities. Finally, we leverage the power of Gated Recurrent Unit (GRU) networks for classification, enabling the model to effectively learn complex relationships within the extracted features and achieve accurate cancer detection. Our proposed ALO-based segmentation approach, incorporating either Convolutional Neural Networks (CNN) or GRU architectures, achieves superior performance compared to existing solutions in breast cancer segmentation. Extensive simulations demonstrate significant improvements in accuracy, precision, recall, and F1-score. Notably, even without utilizing the computationally intensive GRU architecture, our CNNbased model exhibits near-optimal results, making it a valuable option for time-sensitive applications.