Breast Cancer Detection Using Deep Learning and Quantum-Inspired Hyperparameter Optimization
Nusrat Kaniz Khan, Muhammad Nazrul Islam · 2025
Breast cancer remains one of the most prevalent causes of mortality among women that highlights the critical need for accurate and early detection. The objective of this study is to compare the deep learning-based computer-aided diagnosis (CAD) systems for breast cancer detection, to explore the impact of hyperparameter optimization techniques, and to propose an ensemble model with optimized hyperparameter to improve classification accuracy for breast cancer detection. To attain these objectives, a representative subset of mammographic data was used to assess the performance of state-of-the-art pretrained models, including VGG19, ResNet50, MobileNet, In-ception V3, and DenseNet. It has been observed that although the classical tuning methods (e.g., Bayesian optimization) significantly enhanced model accuracy, but incurred high computational costs. Thus, a quantum-inspired optimization strategy based on the Quadratic Unconstrained Binary Optimization (QUBO) formulation was employed here. Despite the reduced data volume and computational load, the QUBO-based approach achieved competitive diagnostic performance. Furthermore, an ensemble model with optimized hyperparameter was introduced to mitigate overfitting and improve generalizability for yielding an accuracy of 92 % with balanced precisions and recalls. The findings firstly underscore the complementary strengths of classical and quantum-inspired tuning, and secondly, their combined application within ensemble frameworks that offer a robust and scalable pathway for enhancing CAD systems in breast cancer screening.