Breast cancer detection using optimized hyperbolic graph attention with bidirectional convolutional neural network

Breast cancer is a prevalent form of cancer affecting women globally, necessitating reliable detection methods to improve survival rates and treatment outcomes. Major existing issues in breast cancer detection include challenges with accuracy, as current methods struggle to reliably differentiate between benign and malignant lesions, potentially resulting in misdiagnoses. To overcome these issues, this research introduces a new method called cloud-based breast cancer detection using optimized Hyperbolic Graph Attention with Bidirectional Convolutional Neural Network (HGABCNN) to identify breast cancer using thermal images. To optimize the performance of the proposed method, the Adaptive Gold Rush optimization technique is integrated. Techniques such as semantic edge-aware median morpho filtering, Swin-based feature extraction and self-adaptive correlation-constrained Fuzzy C-Means clustering are used for preprocessing, feature extraction and segmentation. It improves image analysis accuracy. The proposed method achieves outstanding results, hitting a maximum accuracy (98.6%), recall (98%) F1-score (98.3%), and precision (98.6%) in performance metric comparisons. As a result, this technique outperforms existing methods, showcasing its potential for advancing breast cancer detection and diagnosis.