Ensemble Learning for Breast Cancer Prediction Using FNA Histopathology Images

Abstract

In this chapter, prediction of breast cancer has been carried out from fine needle aspiration (FNA) histopathology images using TensorFlow. The prediction accuracy has been improved by ensembling of neural networks with a custom NN and 2 pre-trained models MobileNet and VGG16. The FNA images are augmented and combined by concatenation, average and weighted average. Performance metrics like accuracy, loss, classification reports, and confusion matrices have been used. When these highly accurate and efficient models are employed for medical purposes, it serves as an invaluable assistance for early identification and detection of breast cancer which is highly advantageous to medical professionals and the society.