Machine Learning Model for Breast Tumor Classification on Histopathological Images

Abstract

Advances in breast cancer screening programs have received significant attention due to their potential to improve early detection, accuracy, and efficiency in diagnosing breast disease. Timely diagnosis and accurate staging are crucial for effective treatment of breast cancer. To systematize them, software detection methods based on machine learning are used. These articles allow breast histopathological images, to be carefully examined, allowing abnormalities (cancer detection) to be diagnosed quickly. Ultimately, this results in increased treatment effectiveness and reduced mortality rates. This methodology innovation allows radiologists to detect malignant tumors without the need for surgical procedures. The change in strategy is mainly due to the widespread adoption of machine learning models and related technologies. A new approach is presented that consists of two critical steps: using Convolutional Neural Networks (CNNs) to extract biological features and then using Support Vector Machines (SVMs) to reliably detect breast tumors and classify them as benign or malignant. However, the traditional CNN-based SVM model has encountered overfitting problems due to the use of a large amount of training data, despite the initial promises. This approach combines CNN, rectified linear unit (ReLU) structure, and SVM using combined features learned from CNN. The success of this tactic is measured using performance metrics that include precision, recall, and accuracy. The SVM-CNN model tested on two state-of-art datasets, i.e BreakHis and Bach. Proposed method achieved 98% accuracy on the BreakHis dataset, indicating its potential to revolutionize breast tumor classification and diagnosis.