Transforming Early Breast Cancer Detection: A Deep Learning Approach Using Convolutional Neural Networks and Advanced Classification Techniques

Breast cancer is a major global health problem, with the WHO estimating approximately 2.3 million new cases every year. In this study, we present a new approach to improve the early detection of breast cancer using deep learning methods and visual inspection of histopathological images. In a world where access to a doctor with specialised knowledge is limited, this study attempts to address the important limitations of current diagnostic strategies that facilitate the efficient detection of diseases. In the proposed method, we use transfer learning with a combination of classifiers, such as SVM, decision trees, and K-nearest neighbours, while implementing two different feature extraction approaches, PCA for dimensionality reduction and no PCA. The approach includes a full evaluation system using metrics such as recall, accuracy, precision, and ROC curves to evaluate the performance of the models. This yields major performance gains for almost all classifiers. The experimental results showed that the SVM classifier with PCA feature extraction obtained the best accuracy of 99.5% with 99.2% precision and 99.6% recall, indicating a significant improvement over the current approach. Even without PCA implementation, the Decision Tree classifier also performed well, scoring 99.4% accuracy. In particular, the application of PCA improved the accuracy of the Boosted Tree from 82.9% to 91.01%. The execution times of classifiers varied significantly; for example, SVM, which is the fastest one as of now, with an execution time without PCA of 38.24 s. This study suggests a potential clinical tool that combines advanced deep learning methods and subsequent classification in real healthcare systems to improve breast cancer detection capabilities. This shows that the high accuracy of this framework, coupled with its computational efficiency, makes it an invaluable tool for real-life clinical applications, which could minimise misdiagnosis and lead to better patient outcomes through earlier detection of respiratory viruses worldwide, especially in remote areas with limited health-care resources.