Rank-based Ensemble of CNN Models for Breast Cancer Detection in Histopathology Images

Abstract

People today suffer from a variety of diseases as a Breast cancer is one of the most common types of cancer in women worldwide. Breast cancer can be invasive or non-invasive. Invasive breast cancer can spread into surrounding tissues also in distant organs whereas non-invasive breast cancer spread surrounding the milk ducts or lobules of the mammary glands in the breast. However early detection followed by the diagnosis can not only reduce the treatment cost but also increase the chance of survival. One of the most effective methods for the detection of tumor malignancies is breast histopathology image analysis. However, manual image analysis for breast histopathology is subjective, time-consuming, and prone to human error. Convolution neural network (CNN) based models have produced encouraging results for the classification of breast histopathology imaging. In this work, we will combine the results of three transfer learning experiments to create a model utilizing a novel rank-based ensemble method rather than depending on a single CNN model, namely GoogleNet, VGG16 and MobileNetV3 Small. Here, it is suggested that the 2-class classification problem for breast histopathology pictures can be handled using the Gompertz function. This strategy yields superior classification results in contrast to publicly available standard datasets such as BreakHis and ICIAR-2018, state-of-the-art methodologies.