AI is here to stay in healthcare

If you’re working in the digital health space, it's hard to avoid robust discussions about artificial intelligence or machine learning. If you look at any programme for any healthtech conference, you’re almost guaranteed to see at least one speaker or presentation on the topic.
‍
In a recent survey, 487 pathologists reported that “the majority of respondents carried generally positive attitudes towards AI, with nearly 75% reporting interest or excitement in AI as a diagnostic tool to facilitate improvements in workflow efficiency and quality assurance in pathology.”

In another survey of specifically ophthalmology, radiology/radiation oncology & dermatology, 71% believed artificial intelligence would improve their field of medicine, and 85% feels that the medical workforce needs would be impacted by the technology within the next decade.

With massive amounts of R&D funnelled into this area, from the ground-breaking fields of oncology and neurology, to now sports-science, ophthalmology as well as more administrative applications like patient reported outcomes & billing/insurance, it’s hard to say that people are still on the fence.
‍

AI in practice

There is a large number of start-ups in this space, including many here in Australia, that are looking to apply Machine learning and AI techniques to leverage efficiencies in image & data processing, or provide insights into population health. 
‍
One start-up that Curve has worked with from ideation through to Minimum Viable Product is Macuject.

‍Conceived by Dr. Devinder Chauhan in 2016, Macuject combines patient management, clinic decision support & a trained model for imaging inferencing, aiming to reduce costs, improve patient outcomes and provide guidance around best treatment practices to help address the massive global problem of undertreatment of macular degeneration. 

When Dr Chauhan approached Curve, he saw the perfect opportunity to revolutionise his field. 

We helped guide him to a minimum viable product with our design principles (human-centered and iterative, to ensure you solve the right problem), through some ups & downs, some beta-testing, more feedback, key opinion leader engagement, and now they’re accelerating with a growing team of their own. 

All that couldn’t come at a better time. In a presentation at EURETINA 2019 (European Society of Retina Specialists), attendees were asked about the impact of AI in their field:

Curve has worked on training models for physical rehabilitation, to ensure adherence of exercises for patients in-home therapies, and is currently working around patient administration and patient reported outcomes. 

We’ve also invested energy into some of the off-the-shelf services now available from AWS, IBM & Google, using their natural language processing to help quickly get information out of the big data healthcare creates.
‍

The dark side of AI in healthcare

While there is an overwhelming feeling that healthcare will be positively impacted and evolved due to AI in many and varied ways, it’s important to be aware of its potential pitfalls and downsides. 

Some research has shown that deep learning has the unfortunate and damaging tendency to perpetuate discrimination. 

Poor AI bias is a nuanced and complex issue. It’s often the result of a confluence of decisions including a poorly framed problem, unrepresentative data, and discriminatory attribute selection when preparing data - which is what makes it difficult to pinpoint and prevent.
‍
Given the existing disparities in pain in underserved populations, AI has the potential to be a driver of change and improvement, as long as these issues aren’t left unaddressed.
‍

The future of AI in healthcare

There are many instances in which AI has the ability to perform healthcare tasks as well as - or in some cases, better than - humans.

But there’s certainly a long road ahead before the large scale automation of healthcare professional jobs is realistic , or even feasible. The “human” factors are so important in this industry, and even with the best natural language processing or chat-bots, our desire to be seen and addressed by a human - not a bot - for connection, will remain a key element of healthcare. 
‍
In the meantime, AI will continue to be implemented in key areas such as diagnosis and treatment recommendations, patient engagement and adherence, and administrative activities, and our role is to remain vigilant and conscious of the ethical issue that AI might also give rise to as we make better and broader use of it. It’s about fit for purpose, and if we can integrate these technologies to reduce errors in repetitive processes or tasks, or allow clinicians to focus on what matters, then we can move to a world of AI-Assisted workplaces that focus on our strengths.