An ML Based Framework for Classification of Breast Histopathological Images Using Fused Feature Set

The integrated approach of medical imaging and machine learning (ML) techniques has shown encouraging results in diagnosis and early detection of the breast cancer. The present work leverages the power of ML techniques in the classification of breast cancer using BreakHis dataset. In proposed methodology Un-sharp masking technique has been used to increase the sharpness of images, followed by extracting the texture and color features from both original as well as pre-processed images. The feature engineering step is followed by feature fusion wherein performance of different combinations of fused feature sets has been analyzed using support vector machine (SVM) classifier. The results obtained against various performance metrics show that highest accuracy of 93.08 % is achieved using fused feature set containing texture features obtained from pre-processed images and color features obtained from original images. The present work represents a significant step towards automating breast cancer diagnosis, offering the potential for early detection and improved patient care.