PSO-optimized fractional order CNNs for enhanced breast cancer detection
Abhinay Yadav, Vaegae Naveen Kumar
Abstract
Breast cancer remains a leading cause of cancer-related deaths in women worldwide; its early and accurate detection will increase survival. The inevitable challenges of this traditional diagnostic approach include relatively high false-positive and false-negative rates when screening dense breast tissue. These unavoidable challenges need to be overcome. In this regard, an effective approach in the area of detection of breast cancer has been presented with advanced filtering image techniques that integrate Fractional Order Convolutional Neural Network (Frac-CNN) and are further optimized with Particle Swarm Optimization (PSO). This would be towards designing a reliable and efficient tool for early breast cancer detection. Adaptive filtering helps reduce noise significantly by using this process of advanced image preprocessing for enhanced image quality. Hyperparameter optimization using PSO ensures that only the best configurations for the model are considered. The fractional calculus adopted by the Frac-CNN architecture is helpful in capturing intricate patterns in mammograms for improved detection accuracy. This proposed work evaluates the method more rigorously compared to state-of-the-art techniques and shows that it reaches very high levels of accuracy, specificity, and sensitivity. The method obtained an accuracy of 99.35%, a specificity of 98.2%, and a sensitivity of 99%, which outperforms traditional methods. This proposed work will improve patient outcomes and treatment strategies by providing a reliable and efficient tool for early breast cancer diagnosis. • An efficient image-filtering method enhances mammogram quality via adaptive mean filtering. • The proposed Frac-CNN effectively captures complex patterns in mammograms. • PSO optimizes Frac-CNN hyperparameters for improved accuracy and efficiency. • The method outperforms state-of-the-art models in accuracy, specificity, and sensitivity. • Results demonstrate the model's reliability for breast cancer detection.