Healthcare as a field isn’t typically associated with fast, game-changing innovation.
The reams of red tape set up to regulate an industry charged with the care of human lives
often result in a notoriety for bureaucratic sluggishness. Getting a drug from the research
lab through clinical trials and governmental approval to the pharmacy can take decades.
Within this context, the spate of so-called “quantum leaps” in healthcare in recent years can come as a bit of a shock. Most of these advancements are powered by AI and boy, do they make great headlines. From early cancer diagnosis and discerning the genetic causes of autism, the general interest reader quickly gets the impression that artificial intelligence in revolutionising the field. But trying to understand progress which comes from all over the place can be disorienting, especially for a technology as new as complex as AI. This article espouses a more big-picture overview of AI in healthcare by reviewing the three main stages of diagnosis, treatment and end-of-life care, as well as the role of research throughout these stages.
One of the earliest uses of AI in healthcare was diagnosis, most notably for various forms of cancer. Traditional methods of detection are often either inaccurate or invasive. According to the American Cancer Society, close to 1 in 2 mammograms yield false results, which increases the need for biopsies in breast cancer diagnosis. The use of AI in the early detection of breast cancer has been reported to be 30 times faster with a 99% accuracy. More recently, Google researchers in the US have successfully developed and trained a deep learning algorithm with vast amounts of CT scan data in the detection of lung cancer, yielding a detection frequency 5 percent higher than the leading experts in the field.
Closer to Asia, doctors in Singapore are using a Chinese-developed AI-powered
smart assistant to recommend treatment for over 1,500 diseases. Diabetes is a particularly
prevalent issue in the tiny city-state, and the AI tool aims to reduce complications of
treatment including stroke and kidney failure. Chinese insurance giant Ping An is bankrolling
the joint venture with public health care group SingHealth, which leaves one guessing whether
it was also behind the assistant’s tongue-in-cheek sobriquet AskBob.
The most striking characteristic about this newly developed tool is its continuous synergy with medical developments in the field. With the help of natural language processing and archive digitalisation, AskBob is consistently analysing troves of medical research literature to better inform the recommendations it provides to experts. This is representative of the use of AI for research purposes in the healthcare industry.
End of life care
A newly opened exhibit at the Science Museum in London poses an ominous question
“Will you be the first person to live a thousand years?” Its optimistic tone belies the
fact that the current generation will die slower than ever before, and in completely different
ways. Conditions like dementia, osteoporosis and heart failure are all overshadowed by the
loneliness underlining the latter stages of one’s life. Its an eerie modern take on The Bucket
The role of AI in this stage is probably the least prominent at this moment, but its clear that robots have immense potential in end of life care. Providing company and helping the elderly remain independent reduces the ever-increasing need for care homes and hospitals which are being stretched to the limit especially in the aging populations of developed countries.
Challenges on the horizon?
Healthcare has an inherent advantage over other industries when it comes to the incorporation of AI as a value-adding technology, namely the sheer volume of data available. It is in fact estimated to surpass finance, media and manufacturing with compound annual growth of 36% through 2025. However, in an insightful editorial over at Information Age, Lauren Maffeo warns that “what healthcare data provides in volume, it lacks in quality.”
"" Without a high amount of quality data to train such systems, there’s no hope of building AI into clinical workflows. "" — Lauren Maffeo
Healthcare visionaries often quote the triple aim initiative for the advancement of healthcare: reduce costs, improve quality of care and patient experience. While it is easy to explain how artificial intelligence might achieve all of these goals, it remains to be seen whether or not it can be implemented as a technology with sufficient scale and speed to upend the industry itself à la the invention of penicillin less than a century ago.