After a few false dawns, artificial intelligence (AI) and machine learning are now at the point where they can start to make a real difference in healthcare.
The power of these two increasingly mainstream technologies lies in their ability to recognise patterns in data much faster and more accurately than humans. This enables the development of ‘smart’ processes that can automate the analysis of clinical information and create ever more streamlined workflows that save significant and potentially crucial time along the diagnostic journey.
So, by using APIs (application program interfaces), for instance, it’s possible to take imaging data directly from a Vendor Neutral Archive (VNA) and apply an appropriate AI algorithm that enables even the inexperienced to effectively and confidently interpret medical images.
But if machine learning and AI systems are to become central components of the healthcare environment, organisations need to make sure that the data they are feeding into them is of a high enough quality. If the data sets being used are neither accurate nor robust, there is the potential for clinical outcomes to be compromised.
Avoiding this scenario involves finding ways to consolidate what is often disparate and totally unstructured data into patient records that are as complete, detailed and accessible as possible.
This is not yet a simple process, given that Electronic Patient Records (EPRs) – now the most significant repository of digital clinical knowledge – are really only able to handle structured clinical information and so can’t capture anything that doesn’t fit snuggly into discrete data fields.
And since around eight out of ten pieces of patient information consist of unstructured data, according to research by Gartner and Merrill Lynch, there is still an information black hole to be filled, with a wealth of knowledge left dispersed across organisations, in systems that leave it either invisible or inaccessible to clinical decision makers when they need it.
Medical images are particularly prone to falling between the cracks, given the lack of interoperability between different systems, with data being stored in proprietary formats that are incompatible with one another.
So, if clinical decision makers are to have any awareness of this potentially insightful data, it needs to be brought together in one place. The obvious solution for achieving this is an Enterprise Content Management (ECM) system with the capability of handling both current and historic patient information from multiple sources and in diverse formats, including paper documents, handwritten notes, emails, faxes and, most particularly, images. Once this is done, AI and machine learning can be employed most effectively to improve and accelerate the diagnostic process, so patients enjoy the best care and experience.
Given its transformational potential, more and more healthcare organisations are now declaring an intention to increase their investment in artificial intelligence and machine learning systems.
Of course, there will still be very much a central role for the human element, simply because this new technology doesn’t have any ‘common sense’ when it comes to making clinical judgements and so needs to be overseen by healthcare professionals. In other words, the widespread introduction of AI and machine learning won’t mean that we are about to enter some strange era of hands-off medicine, but one where clinicians are made more effective thanks to these advances – and thus able to deliver the best care possible.
It will be some time before we see real benefits coming from these technologies, but they are already making a difference. Dermatologists, for instance, are already detecting skin cancer with much greater accuracy and effectiveness because of the AI systems they are now using.
These are exciting times for artificial intelligence in healthcare and will become ever more so as it is applied to tackle more and more of the tedious and time-consuming processes that lie behind so much of modern-day medicine.
And although healthcare may have been somewhat late coming to the party compared with some other sectors, we are now at the stage where the perceived potential of AI is being translated into practical reality. So, we can be pretty sure that when it comes to improving patient care and increasing the productivity of healthcare professionals, this is going to be genuinely game-changing.
So how transformative will AI be in healthcare?
When a report from the McKinsey Global Institute estimates that the overall impact of these two technologies will be 3,000 times greater than that of the Industrial Revolution, we can only imagine.
Saduf Ali-Drakesmith is a diagnostic radiographer and imaging expert with Hyland.
