| Summary: | Breast cancer is predominantly seen in women and is the leading cause of death in females worldwide. Diagnosis of breast cancer using biopsy tissue images is expensive, time-intensive, and fraught with conflicts among doctors. Pathologists can now diagnose breast cancer more consistently and promptly because of advances in the Computer-Aided Diagnosis (CAD) system. As a result, there has been a surge in demand for CAD-based machine learning techniques. This study describes a “BreastMultiNet” framework that focuses on the transfer learning concept for identifying distinct types of breast cancer by utilizing two publicly available datasets. The suggested “BreastMultiNet” architecture allows rapid and comprehensive breast cancer diagnosis. The suggested scheme extracts features from microscope images with the help of well-known conventional and deep learning models such as HOG, LBP, SURP, DenseNet201, and VGG19. Comparatively, transfer learning models provide good accuracy than conventional models. The collected properties of transfer learning models are subsequently dispatched into the summing layer, resulting in a fused vector. The proposed framework achieves 99% and 95% classification accuracy on both BreakHis and ICIAR dataset respectively, outperforming all the other state of the art techniques. In terms of accuracy, the “BreastMultiNet” framework may be employed as a modeling approach in hospitals and medical care contexts.
|
|---|