Artificial Intelligence for Medical Diagnosis
Jonathan G. Richens, Albert Buchard
Abstract
Medical diagnosis has been one of the primary targets of Artificial Intelligence research since the inception of the field. In recent years, rapid advances in Artificial Intelligence have seen the emergence of diagnostic algorithms that perform as well as clinicians and can be applied at scale in clinical practice. This chapter presents a broad picture of the foundations, history, and the current state of AI in medical diagnosis. We provide an overview of the complex and interdependent tasks required to perform diagnosis and explore how ideas from the study of diagnostic reasoning and diagnostic errors can guide the effective development and deployment of Artificial Intelligence solutions. We then review the three main approaches to diagnostic AI; rules-based, model-based, and machine learning, detailing their strengths and weaknesses, and how each of these approaches tackles diagnosis from a different angle.