Advanced Deep Learning for Medical Image Analysis

Abstract

The application of deep learning is evolving, including in expert systems for healthcare, such as disease classification. Several challenges in the use of deep-learning algorithms in application to disease classification. The study aims to improve classification to address the problem. The thesis proposes a cost-sensitive imbalance training algorithm to address an unequal number of training examples, a two-stage Bayesian optimisation training algorithm and a dual-branch network to train a one-class classification scheme, further improving classification performance.