Automated breast cancer diagnosis using a spherical convolutional cross-contextual attention network with mountaineering team-based optimization

Breast cancer diagnosis remains challenging due to limitations in existing imaging and diagnostic techniques, such as reduced high false-positive rates and sensitivity in thick breast tissues rates, and inter-observer variability, which can lead to inaccurate or delayed detection. Early and accurate diagnosis is therefore critical for enhancing treatment results and survival rates. In addition, existing artificial intelligence and deep learning models often face challenges in capturing global spatial relationships, contextual feature interactions, and robust parameter optimization, limiting their diagnostic performance. To solve these problems, this paper proposes an Spherical Convolutional with Cross Contextual Network with Mountaineering Team-Based Optimization (SC3AN-MTBO) solution to boost the accuracy of breast cancer diagnosis. Initially, mammography images are collected from the MIAS and CBIS-DDSAM data sets. Then, to enhance the image, the Adaptive Side Window Joint Bilateral Filter (ASWJBF) is used as a preprocessing step, which preserves edges and removes noises in breast cancer images. Finally, to segment malignant tumors, Convolution Transformer (CT) is employed. Then, Clifford Fourier Mellin Transform (CFMT) improves invariant feature extraction in breast cancer diagnosis. After that, classification used for Spherical Convolutional with Cross Contextual Attention Network (SC3AN) improves the accuracy of breast cancer diagnosis. Finally, Mountaineering Team-Based Optimization (MTBO) improves model parameters to increase classification accuracy. The suggested SC3AN-MTBO approach offers high precision levels in detecting breast cancer by scoring very high on accuracy (99.8%), F1-Score (97.4%) on MIAS database, high on accuracy (99.7%), F1-Score (98.6%) on CBIS-DDSAM database, and an error rate of 0.1%.