RESEARCH & DEVELOPMENT BIOINSPIRED ELECTRONICS
– diffusion and drift. A human cell
doesn’t have electrons, but it has got
ions which obey the same physical
properties.”
Georgiou explains that, because
of these parallels, electronic systems
can be architected using transistors
that obey the same physical laws as
biological cells. This approach was
originally pioneered in the early 90s
by Carver Mead at The California
Institute of Technology.
Mead established neuromorphic
electronics – a new area of thought
where he drew analogies between
transistor physics and neural cells
in the brain – and came up with the
hypothesis that because the laws
which govern both are the same,
you should be able to architect
systems which are similar to the
brain. Georgiou says neuromorphic
electronics are arguably the
foundation for arti cial intelligence
and machine learning.
Extending neuromorphic electronics
“About 15 years ago, I thought I’d
take that approach a step further and
ask, ‘can we do something similar
with metabolic systems?’ I wanted
to see if we could create an arti cial
pancreas for the management of
diabetes and looked at the biological
cells in the pancreas responsible for
glucose management, called beta
cells. I saw that there were analogies
between the way those cells behaved
and CMOS transistors.”
In the human body, blood glucose
levels are regulated by the gradual
release of the hormone insulin. Type
1 diabetes is an autoimmune disease
where a person’s immune system
kills off the beta cells that ordinarily
create insulin.
Based on the thinking that
electronic systems can be architected
to mirror biology, Georgiou developed
a ‘silicon beta cell’ as an arti cial
replacement for the biological beta
cell. He then used that silicon beta
cell as the chip within the bio-inspired
arti cial pancreas, a fully closed
loop system designed to act as a
substitute to the biological pancreas
for the treatment of diabetes.
This research was originally
Georgiou’s PhD project and, since
then, it’s gone on to grow as an
initiative in his lab. Georgiou says,
“We’ve now clinically trialed the
device on 65 patients and I’m
pleased to say, as of last December,
we have sent the device home
fully unsupervised in a free-living
environment.”
He explains that the trials that
had been carried out prior to this
were within hospital environments,
and the closed loop system has been
shown to manage diabetes. Now, a
new direction in his research lab is to
bring in wearables to try and monitor
additional parameters.
Stress and exercise are important
factors in managing diabetes, so
being able to detect these and feed
those into the arti cial pancreas
system will hugely improve the
treatment outcome. The arti cial
pancreas system itself uses an
insulin pump to replace insulin
injections.
“The problem with insulin
injections is that you still need to
monitor your blood glucose through
nger prick samples. Then, to gure
out how much insulin you need,
you need to do some mental
arithmetic.
“What we offer is a fully
closed loop system, which
works in real-time. Every ve
minutes it senses glucose and
calculates how much insulin
you need. The insulin which is
calculated is based our beta cell
model so it’s physiological, and
then it sends that information to
the insulin pump, which then pumps
in the insulin continuously.”
Improving treatment outcomes
This approach, even in this trial
phase, is already showing patients
can have better treatment outcomes
than existing methods. Georgiou
says, “There are two bene ts. One
is, for the rst time, you’re providing
patients with a fully automated way
to manage their diabetes. They don’t
have to inject three times a day or
do nger prick samples because
they’ve got continuous monitoring
and infusion.
“The second, we’ve done a study
where we compared treatments and
we’ve shown that patients spend a
lot more time in the healthy blood
glucose range, between four and
eight mmol/l, that you would expect
for a healthy individual.
“We’ve also reduced the number
of hyperglycemia events that patients
experience.
“Hyperglycemia is very dangerous,
because when you infuse too much
insulin and the body is depleted of
glucose, you end up fainting and
it can be fatal. We showed in a
head-to-head study that we totally
eradicated hypoglycemia overnight.
And compared to standard therapy,
we actually reduced it by as much as
20%.”
With these results, bio-inspired
electronics should clearly be taken
seriously as a way to transform
outcomes for patients with various
illnesses or diseases. This research
project might be focused on diabetes,
but Georgiou and his colleagues
at Imperial College London have
developed electronic chips
that mirror biology to help
diagnose and treat numerous
conditions. Those include
a microchip-based for early
detection and therapeutic
monitoring of breast
cancer and a chip to better
understand of the role of
hormones in appetite regulation
and their impact on conditions such
as obesity.
Electronics are already being used
to carry out bodily functions, but for
this to be automated, they need to
be architected in a way that mirrors
biology.
It’ll be interesting to see what
other chips are built to replicate the
extreme accuracy, precision and
robustness of human organs.
“The human brain
is seven orders of
magnitude more
effi cient than
the world’s best
microprocessor
and, if you look at
the components it
uses, they’re not
entirely precise.”
Dr Pantelis
Georgiou
Above: An example
of a chip developed
to mirror a biological
function for neural
recording
www.newelectronics.co.uk 25 February 2020 21
/www.newelectronics.co.uk