Detection of Metastasis in Breast Cancer Patients using Deep Neural Networks by Ensemble Stacking

Abstract

Because of metastasis, which spreads breast cancer from one region of the body to another through the lymphatic system and bloodstream, breast cancer has become a severe issue for women, and early detection of metastasis in breast cancer patients is essential to lowering the death rate of female patients. The challenges faced by transfer learning are Choosing the Right Pre-Trained Model, Overfitting, and bias. Subsequently, utilizing histopathological images, a stacking ensemble approach is used by considering the pre-trained deep convolutional neural networks VGG-19, InceptionResNetV2, and the newly designed architecture called the Baseline model. These models were trained separately on the dataset before being integrated using the stacking ensemble technique. Among the pre-trained models, the baseline model got better accuracy. When these three models are combined using a stacking ensemble the ensembled model obtained better results than individual pre-trained models.