Three Ways Artificial Intelligence Is Changing Healthcare – And Six Principles To Ensure Its Success
Dr. Sina Habibi, Co-founder & CEO Cognetivity Neurosciences.
Artificial Intelligence (AI) has been used to revolutionize healthcare for some time, but all the signs suggest that AI is poised to completely take the medicinal world by storm over the next few years.
Governments — from the U.K. and the U.S. to Germany, Qatar and China — have published strategy papers on AI in healthcare and backed up their words with multimillion-dollar funding. The WHO recently issued its first global report on AI in healthcare. And global equity funding in AI health start-ups has risen quarter on quarter since the end of 2019, reaching a record $2.5 billion in the first quarter of 2021.
All this buzz is a reflection of the enormous range of healthcare-related activities in which AI can have a big impact. In this post, I'm going to highlight three areas where this is happening and the guiding principles that will enable us to ensure AI changes healthcare for the better, and do so as quickly as possible.
Diagnostics is the first area many people think about when it comes to AI in healthcare, largely due to the high-profile work of Google's subsidiary DeepMind. In 2018, DeepMind announced that it had used deep learning to create software capable of analysing 3D retinal scans and detecting over 50 eye diseases almost as accurately as specialist doctors at the world-leading Moorfields Eye Hospital.
In 2020, they created an early warning system for age-related macular degeneration (AMD) — the most common cause of blindness in the developed world. Trained on a dataset of retinal scans, the system performed as well as, and in some cases better than, expert clinicians in predicting whether patients would develop “wet” AMD, a particularly dangerous form of the disease, in the next six months.
It's clear such programs have massive potential to drive early diagnosis and treatment, enhance the efficiency of patient care and improve clinicians' working lives. While DeepMind's solutions have not come to the clinic yet, dozens of AI-powered tools for detection and diagnosis have already been FDA approved for a wide variety of cases, including cardiac arrhythmias, diabetic retinopathy, early-stage breast cancer and strokes.
Beyond cutting-edge diagnostics, AI is also revolutionizing the administrative and operational aspects of healthcare.
These may be less glamorous areas — but that's exactly what makes this type of transformation such a tantalizing prospect. Physicians don't get into medicine to do desk work, yet a 2016 study found that they spend almost twice as much time on it as they do interacting with patients.
A valuable solution to this problem has been to harness the vast improvements in natural language processing (NLP) that have powered the commercial success of products such as Amazon's Alexa. AI-based medical dictation software can enable automated note-taking and data entry into electronic health records. One firm that does this, Suki, claims to help doctors finish their notes 76% faster on average.
Another shortcoming in healthcare is poorly managed patient flow in hospital wards, whereby issues such as inefficient surgical scheduling and disorganized patient transfers can contribute to overcrowding and adverse patient outcomes.
Qventus is an AI-based software platform that orchestrates patient admissions and discharges to solve such operational difficulties. Deployed in hundreds of wards across the U.S., their technology reportedly enables emergency departments to achieve a 20% reduction in the time taken for patients to see a doctor and a 15% reduction in the length of patients' stays.
Third, there's drug discovery and development. Given the billions of dollars it costs to bring the average drug to market, the pharmaceutical industry has turned to AI to enhance virtually every aspect of the entire process.
According to a paper published this year, AI has now found favour within drug design and screening, product development, manufacturing, quality assurance and control and even product management. It is also being used to enhance the clinical trial process, which generally takes around ten years and sees nine in ten candidate drugs fail.
One of the biggest challenges for researchers is to recruit sufficient numbers of suitable trial participants. This is expensive and time-consuming, and recruitment targets are rarely met, leading to underpowered results.
To tackle this, the start-up Trials.ai uses NLP and machine learning to mine vast quantities of trial-related documents such as drug labels and published papers. It analyses the implications of a trial's proposed eligibility criteria, outcome measurements and assessment schedules, among other factors, for key outputs such as trial duration, cost and patient burden. Hence, it enables researchers to optimize clinical trials, resulting in more efficient drug development.
Other solutions have taken different approaches. Software developed by AiCure, for example, uses an AI-based facial recognition and patient engagement platform to monitor medication intake and has been shown to significantly increase adherence rates in clinical trials.
Driving Positive Change
Exciting as these case studies are, it's important to avoid the pitfalls in the implementation of such powerful technology.
The WHO mentioned earlier focuses on these pitfalls and offers six principles for maximizing the benefit of AI in healthcare: 1. Protect human autonomy, 2. Promote well-being, 3. Ensure transparency and explainability, 4. Foster responsibility and accountability, 5. Ensure inclusiveness and equity, and 6. Promote AI that is responsive and sustainable.
The report is overwhelmingly positive about the promise of AI in healthcare, but these principles are undeniably valuable. What's more, for those working at the intersection of AI and medicine, they should be seen not just as moral responsibilities but sound business practice.
Firms that make their AI models easy for clinicians to understand, demonstrably respect patients' privacy and build quality improvement into their products will achieve greater uptake and commercial success. Thus, these principles can guide us to both ensure more favourable patient outcomes and achieve faster uptake and healthcare improvements.
I welcome the pace of technological change we are seeing in AI, and I also welcome guiding principles that can help us maximize its benefits. In light of both, the future of healthcare certainly looks bright.