Machine Learning for Healthcare: Introduction

Abstract

Machine Learning (ML) is an evolving area of research with lot many opportunities to explore. “It is the defining technology of this decade, though its impact on healthcare has been meagre”—says James Collin at MIT. Many of the ML industry’s young start-ups are knuckling down significant portions of their efforts to healthcare. Google has developed a machine learning algorithm to help identify cancerous tumours on mammograms. Stanford is using a Deep Learning algorithm to identify skin cancer. US healthcare system generates approximately one trillion GB of data annually. Different academic researchers have come up with different number of features and clinical researchers with different risk factors for identification of chronic diseases. More data means more knowledge for the machine to learn, but these large number of features require large number of samples for enhanced accuracy. Hence, it would be better if machines could extract medically high-risk factors. Accuracy is enhanced if data is pre-processed in form of Exploratory Data Analysis and Feature Engineering. Multiclass classification would be able to assess different risk level of disease for a patient. In healthcare, correctly identifying percentage of sick people (Sensitivity) is of priority than correctly identifying percentage of healthy people (Specificity), thus research should happen to increase the sensitivity of algorithms. This chapter presents an introduction to one of the most challenging and emerging application of ML i.e. Healthcare. Patients will always need the caring and compassionate relationship with the people who deliver care. Machine Learning will not eliminate this, but will become tools that clinicians use to improve ongoing care.