The use of Artificial Intelligence in clinical diagnostics: Challenges to consider for implementation

Abstract

Whilst many technological advancements have revolutionised healthcare throughout the 21st century, one of the most significant is Artificial Intelligence (AI). AI is generally regarded as the capability to imitate intelligent human behaviour using machines, and is based on computer science, statistics, algorithms, machine learning, information retrieval, and data science1. AI has permeated into many domains of healthcare including Clinical Diagnostics. While AI chatbots (such as those used in Babylon and Ada) are being used by patients to identify symptoms and recommend further actions in community and primary care settings, more recent advances in the technology with larger datasets have provided users access to a more extensive array of clinical conditions2. However, as these tools are constantly being developed with an ever-increasing dataset of clinical cases, certain challenges threaten the implementation of an accurate and effective model. In this article, the issue of Data Bias, and Data Handling will be examined within the context of Clinical Diagnostics, and how these factors threaten the development of such AI Healthcare tools.