Prof. Mohamed Medhat Gaber (Birmingham City University): Abstract
In this presentation, we will delve into four significant milestones in the development of class decomposition in medical diagnosis. Each milestone represents a crucial advancement in demonstrating the potential benefits of utilising class decomposition for medical diagnosis problems.
By grouping instances with similar feature values into subclasses or clusters, class decomposition consistently enhances the predictive performance of classification models. Over the years, the application of class decomposition has evolved from shallow learning models to deep learning models, encompassing both tabular and image data. This progression highlights the pivotal role of class decomposition in the workflow of medical classification problems solved through machine learning. Through this talk, we aim to shed light on the transformative impact of class decomposition in the field of medical diagnosis. By exploring these milestones, we hope to provide a comprehensive understanding of how class decomposition can revolutionise the effectiveness of medical classification models.