Skip to main content

Blogs • General

What are the prospects for artificial intelligence in healthcare?

Published 27 February 2018

New report moves beyond the hype and speculation.

Artificial intelligence is in the news a lot these days. New machine learning techniques have been unexpectedly successful, despite being poorly understood, resulting in machine domination of the game of Go and prompting speculation about mass unemployment in coming years as a result of AI encroachment on traditional jobs.

Healthcare is already being impacted by AI and machine learning technologies. Applications include enhancing machine vision to detect intracranial brain bleeds, an advanced cognitive assistant to support clinical decision making in radiology and cardiology, an AI platform to analyse genomic data for the development of next-generation medicines, and machine-learning based analyses of online behaviour to enhance cybersecurity.

Healthcare planners clearly need to know more about the likely impacts of AI on healthcare over the next several years. But mere speculation is inadequate to this task: we need careful, well-informed analysis. Enter JASON.

JASON is an independent scientific group that has been advising the US Federal government on scientific and technological issues for over 50 years. In December 2017 they delivered a report entitled Artificial Intelligence for Health and Health Care that focuses on the technical capabilities, limitations, and applications of AI to healthcare within the next ten years.

Some of the themes that emerged from this study are that massive health data sets will emerge, in some cases generated by and feeding back into various “smart” devices, such as a suitably configured smart phone or the Australian-developed "smart stethoscope". These data sets have great potential to support machine-learning systems, but the quality of any such system is only as good as the quality, the quantity, and the breadth of the underlying data. Rigorous curation of data will be needed, as well as equally rigorous validation procedures for AI-based diagnostics.

Privacy considerations will be particularly pronounced in such developments. Transparent operations will be a necessity, as well as active campaigning to guard against the spread of misinformation.

Despite the depth and breadth of the subject matter of this report, it is surprisingly accessible for the non-specialist reader. If you have an interest in the future of healthcare, put it on your reading list.

Share this article