BreastCancerNet: Flask-Enabled Attention-Driven Hybrid Dual DNN Framework for Real-Time Breast Cancer Prediction.

Abstract

Breast cancer is the most prevalent cancer among women and poses a significant global health challenge due to its association with uncontrolled cell proliferation. Artificial intelligence (AI) integration into medical practice has shown promise in boosting diagnosis accuracy and treatment protocol optimisation, thus contributing to improved survival rates globally. This paper presents a comprehensive analysis utilizing the Wisconsin Breast Cancer dataset, comprising data from 569 patients and 30 attributes. We propose BreastCancerNet, a hybrid AI architecture that leverages dual deep neural networks (DNNs) coupled with an attention mechanism to enhance breast cancer diagnosis. The proposed framework integrates two distinct DNNs (DNN-I and DNN-II) to extract diverse feature representations from the dataset, which are then concatenated for comprehensive analysis. An attention mechanism is employed to prioritize critical features, thereby improving the model's focus on essential characteristics of the input data. The final classification is performed using a support vector machine (SVM), achieving an impressive accuracy rate of 99.42% in differentiating between malignant and benign cases. Furthermore, we introduce a user-centric web application that facilitates real-time breast cancer detection by allowing users to input new attributes. This intuitive web interface fosters interactive engagement with the predictive algorithm, potentially enhancing breast cancer screening and treatment outcomes.