New Scientist - USA (2019-11-30)

30 November 2019 | New Scientist | 11

IN A dimly lit room filled with
computers at St Bartholomew’s
Hospital in London, doctors pore
over scans of people’s hearts. Until
recently, medical staff here had to
interpret the blotchy on-screen
images purely by sight. Now
artificial intelligence is helping to
explain what they are looking at.
Charlotte Manisty, a
consultant cardiologist at
St Bartholomew’s, analyses an
MRI scan of a struggling heart
and points to blue smudges over
one area of muscle. The image on
her screen has been coloured in
by AI. A swathe of blue around
the left ventricle, the heart’s main
pumping chamber, means that
not enough blood is getting
to that part of the muscle. The
volume of blood reaching each
bit of the heart is a good indicator
of how well it is functioning.
The AI provides a numerical
estimate of blood flow for each

region too. Previously, doctors

had to eyeball black and grey

scans to make a judgement about

how much blood was present.

Getting an actual number needed

specialists and took several

hours or days.

“All of the things that we’re

working on here are to try to

reduce the training required,”

says Manisty. The AI works

completely automatically and

delivers its analysis in around

2 minutes, she says.

The same system is now

used at more than 30 hospitals

worldwide and has analysed more

than 20,000 MRI scans to date. It

was developed by Peter Kellman

and Hui Xue at the National

Institutes of Health in Maryland

and their colleagues.

To get the algorithm to correctly

identify each bit of the heart in

MRI scans, the team trained it on

more than 1900 scans of around

1000 patients. The system was

then tested against 200 scans from

105 patients to show that it could

reliably select each area of heart

muscle. It proved to be at least

90 per cent accurate in each case.

The system was also previously

trained to quantify blood flow and

compared against cardiac positron

emission tomography, where

it was found to be 92 per cent in

agreement with that method.

Kellman and Xue’s team

plans to upgrade the AI soon so

that it can determine a patient’s

condition, for instance by stating

whether it thinks it has spotted a

blocked artery, diseased tissue or

a healthy heart. Along with other

algorithms that Kellman and Xue

have developed, such as one

that allows scans to continue

even when patients accidentally

move, AI has improved efficiency

at St Bartholomew’s, says Manisty.

Previously, the department

scanned around 25 people a

day – now that number is well

into the 30s.

Shehab Anwer at University

Heart Center Zurich in Switzerland

questions whether the colour

coding could obfuscate certain

features of a heart scan, perhaps

meaning that doctors miss other

signs of disease. Manisty says that

the original, grey scans are all still

accessible in the system.

William Bradlow at Queen

Elizabeth Hospital Birmingham

in the UK says that there is

little risk of distorting the scans.

Interpreting MRI images of hearts

is tricky, but with help from AI,

more doctors could be doing it

on a regular basis, he says.  ❚

NA

SA

/JP

L-C

AL

TEC

H/S

WR

I/M

SS

S/K

EV

IN^

M.^

GIL

L/C

C^ B

Y^3

.^0

20,

The number of MRI scans

the AI has analysed so far

Gege Li

Jupiter’s Great Red

Spot is healthier

than it looks

Solar system

JUPITER’S giant storm, the Great
Red Spot, may not be dying any
time soon. It seems to have been
unravelling for decades, but this
is probably down to the movement
and shredding of clouds rather than
a sign that the storm is abating.
Concerns have been mounting
that the Great Red Spot might
disappear. Once it was big enough
for almost three Earth-sized planets
to fit inside it – now it can hold little
more than one.
Although we know that the storm
has been shrinking since 1878, the
pace of this seems to have picked
up since 2012.
What’s more, photos of Jupiter

taken earlier this year by the

spacecraft Juno showed red “flakes”

measuring 100,000 kilometres

across apparently breaking off

from the Great Red Spot.

But this flaking isn’t actually

a sign that the storm is fragmenting

and dying, says Philip Marcus at the

University of California, Berkeley.

Using computer models, he and

his colleagues found that the flaking

captured by Juno was in fact the

result of rare events: cyclones that

are common in Jupiter’s atmosphere

colliding with lumps of cloud that

hadn’t yet been pulled into the

storm as they passed by.

The impacts broke apart the

clouds, which appear red because

they sit above the Great Red Spot

and are therefore exposed to more

of the sun’s UV radiation. This gives

the impression that parts of the

storm are disintegrating.

Marcus presented these findings

at a meeting of the American

Physical Society in Seattle this

week. He says he was surprised

by how straightforward it was

to simulate the flaking, which

“cried out for explanation”.

“It’s wonderful to see serious

attempts at numerical simulations

being brought to bear on this

complex topic,” says Leigh Fletcher

at the University of Leicester, UK. ❚

The giant storm on

Jupiter may not be

abating after all

Machine learning

Chris Baraniuk

Artificial intelligence is analysing

heart scans in dozens of hospitals