Hospitals
have begun using machine learning to help analyze and collect images, and the
medical applications are endless.
In
2023, radiologists in hospitals around the world will increasingly use medical
images—which include x-rays and CT, MRI, and PET scans—that have been first
read and evaluated by AI machines. Gastroenterologists will also be relying on
machine vision during colonoscopies and endoscopies to pick up polyps that
would otherwise be missed. This progress has been made possible by the
extensive validation of “machine eyes”—deep neural networks trained with
hundreds of thousands of images that can accurately pick up things human
experts can’t.
One
of the most exciting new capabilities of AI is to instruct untrained and
uninitiated people to acquire medical-grade images through a smartphone.
Someone without any medical knowledge will be able to pop an ultrasound transducer
into a smartphone’s base and, with a little gel on its tip, instantly acquire
high-quality images. The AI algorithm instructs the person to move the
transducer up or down, clock- or counterclockwise, and it will automatically
capture the image when it meets the objective standard. This will extend the
ability to perform medical imaging of most parts of the body (except the
brain), anywhere, anytime, and by anyone. Concurrently, algorithms are also
being developed for automated accurate interpretations. In 2023, we will see
more of this in remote parts of the world, perhaps best exemplifying the
potential for AI to reduce health inequities.
The
same deep-learning democratization is progressively taking hold for patients as
well, who can already be notified by their smartwatch’s algorithm that they
have an abnormal heart rhythm (such as atrial fibrillation). In 2023, this will
extend to preliminary diagnosis of all skin lesions, urinary tract infections,
children’s ear infections, and an increasing number of common conditions that are
not life-threatening.
These
are the early steps towards a virtual health coach to ideally prevent
conditions that a person is at increased risk for manifesting, which in 2023
will be used for managing specific conditions such as diabetes, hypertension,
or even depression, with the help of chatbots and human coaches in the
background when necessary.
In
2023, clinicians will also be aided by AI in their daily tasks—particularly by
being liberated from the job of painstakingly typing medical data into the
computer. This burden not only contributes to burnout among physicians, but
markedly detracts from the patient interactions. Natural language processing
and machine learning now enable synthetic notes to be created automatically
from the conversation between doctors and patients at the visit or bedside.
We
have seen the beginning of use of AI for remote monitoring, which is already
preempting the need for hospitalization for patients with Covid-19 by real-time
data capture from wearable sensors. That will only increase in 2023. We still
need more validation trials to show that algorithms can accurately anticipate
early signs of clinical deterioration and intervene, but the implication for
avoiding a large proportion of hospital stays looms large.
Nevertheless,
there remains a dire need to reduce bias and promote privacy and security in
the application of medical AI. Privacy AI computing is starting to take off
with the use of federated and swarm learning, as well as with the increasing
application of edge computing, which uses algorithms fully operating on the
smartphone. In 2023, these strategies will be explored more fully, in a
much-needed effort to not only fully investigate the potential for AI in health
and medicine but also to address its potential flaws and pitfalls.
