Introduction

Artificial intelligence (AI) is conventionally defined as the ability of a computer system to perform tasks that are usually thought to require human intelligence, including reasoning, self-correction and learning.1 Machine learning (ML) is a statistical subset of AI describing the ability of computers to learn to perform tasks without being explicitly programmed to do so. AI and ML rely heavily on appropriate datasets to train algorithms, and digitalisation of health data, coupled with accelerated development of AI methodologies, has led to a surge in investment and development in recent years. Fundamentally, the use of AI allows clinicians to gain data-driven insights into complex clinical associations that would be infeasible to derive from traditional statistical analyses.2

Investment in AI has been increasing exponentially over the past decade. In 2020 alone, global investment in AI start-ups was estimated to be US$67.9 billion, a 40% increase from the previous year.3 Healthcare as an industry has been particularly rapid to foster AI research, given the abundance of available health data, particularly in data-rich fields such as haematology. AI has the potential to perform simple tasks faster, more reliably and more efficiently than humans without being constrained by working hours.4 AI could enhance both the quality of healthcare at an individual level and democratise access to medical services in low-resource settings. This paper will explore how AI could impact three key components of the haematology patient pathway: diagnosis, monitoring and treatment.