Better medicine through machine learning: What’s real, and what’s artificial?
Suchi Saria, Atul Janardhan Butte, Aziz Sheikh · PLoS Medicine · 2018
Artificial intelligence (AI) as a field emerged in the 1960s when practitioners across the engineering and cognitive sciences began to study how to develop computational technologies that, like people, can perform tasks such as sensing, learning, reasoning, and taking action.Early AI systems relied heavily on expert-derived rules for replicating how people would approach these tasks.Machine learning (ML), a subfield of AI, emerged as research began to leverage numerical techniques integrating principles from computing, optimization, and statistics to automatically "learn" programs for performing these tasks by processing data: hence the recent interest in "big data."Although progress in AI has been uneven, significant advances in the present decade have led to a proliferation of technologies that substantially impact our everyday lives: computer vision and planning are driving the gaming and transportation industries; speech processing is making conversational applications practical on our phones; and natural language processing, knowledge representation, and reasoning have enabled a machine to beat the Jeopardy and Go champions and are bringing new power to web searches [1].Simultaneously, however, advertising hyperbole has led to skepticism and misunderstanding of what is and is not possible with ML [2,3].Here, we aim to provide an accessible, scientifically and technologically accurate portrayal of the current state of ML (often referred to as AI in medical literature) in health and medicine and its potential, using examples of recent research-some from PLOS Medicine's November 2018 Special Issue on Machine Learning in Health and Biomedicine, for which we served as guest editors.We have selected studies that illustrate different ways in which ML may be used and their potential for near-term translational impact. ML-assisted diagnosisOf the myriad opportunities for use of ML in clinical practice, medical imaging workflows are most likely to be impacted in the near term.ML-driven algorithms that automatically process 2-or 3-dimensional image scans to identify clinical signs (e.g., tumors or lesions) or determine likely diagnoses have been published, and some are progressing through regulatory steps toward the market.Many of these use deep learning, a form of ML based on layered representations of variables, referred to as neural networks.To understand how deep learning methods leverage image data to perform recognition tasks, imagine you are entering a dark room and