Breast Cancer Severity Classification; Mesh Denoising Graph Triple Attention with Stereoscopic Scalable Quantum Convolutional Neural Networks

Breast cancer diagnosis and classification are often hindered by challenges such as low image quality, overlapping tissue features, and the lack of distinct tumor boundaries. These limitations contribute to high false-negative and false-positive rates, reducing diagnostic accuracy. Additionally, the scarcity of annotated data and the computational complexity of existing models further impact their performance. To address these issues, advanced imaging techniques, improved feature extraction, and robust classification strategies are essential. This study presents and validates the Mesh Denoising Graph Triple Attention with Stereoscopic Scalable Quantum Convolutional Neural Networks (MD-GTA-SSQCNNet) for breast cancer severity classification using the MIAS (Mammographic Image Analysis Society) dataset. The process begins with Adaptive Mesh Denoising Guided Normal Filtering (AMDGNF), which enhances image quality. This is followed by segmentation of tumor-affected regions using the Geometric Algebra Transformer (GAT). Feature extraction and classification are carried out using the Graph Triple Attention with Stereoscopic Scalable Quantum Convolutional Neural Network (GTASSQCNNet). To further improve classification accuracy across multiple breast cancer classes, the Osprey Optimization Algorithm (OsOA) is employed; optimizing the model based on tumor class distribution and affected regions. The Pythonbased implementation using the MIAS dataset demonstrates that MD-GTA-SSQCNNet significantly outperforms conventional approaches, achieving 99.9% classification efficiency and 99.8 % sensitivity. These results highlight the potential for highly accurate automated systems to assist or even replace manual diagnosis in breast cancer detection.