¿ìè¶ÌÊÓƵ

It sounds like the wildest science fiction. An American GI is blown
up on some remote battlefield. As the soldier lies wounded and barely conscious,
a wristwatch-sized monitor transmits details of the person’s physical state
and location back to base. Minutes later, a surgical robot is dropped off
by helicopter and, via a remote virtual reality link, army surgeons hundreds
of miles away patch up the soldier’s damaged body.

This may all seem a little far-fetched, but Richard Satava, director
of the Pentagon’s Advanced Biomedical Technology Initiative, has already
demonstrated that such remote surgery is possible by cutting and sewing
up a pig’s intestine using a remote ‘telepresence’ surgery system . Satava
is now seeking funds to develop a system that can be used on humans.

Such is the impact of computer technology on medicine that for enthusiasts
like Satava, little seems impossible. The technology to support virtual
reality operations already exists, medical robots and remote controlled
probes are being developed, and rapid improvements in computer graphics
are allowing doctors to work with 3D and even animated images that have
more in common with TV graphics than the fuzzy 2D output associated with
traditional medical scanning devices. ‘The information revolution in medicine
is happening now. This moment in time is seeing the greatest revolution
in medicine,’ says Satava.

Well, yes . . . perhaps, replies Richard Kitney, director of biomedical
systems research at London’s Imperial College. Kitney believes there are
still a few problems that need solving before medicine reaps the full benefit
of these emerging technologies – the biggest of these is the absence of
standards. If techniques such as transglobal remote surgery are going to
be put into practice, the medical profession will need a standard format
for data storage and transmission to ensure all the different imaging devices,
electronic instruments and computerised information systems can communicate
with each other.

‘At the moment, it’s hard enough to get an image off a medical scanner
and across the room to a (computer graphics) workstation, let alone across
the world,’ says Kitney. ‘Each manufacturer uses their own data format,
so there is no compatibility. Standards for networking medical images are
not even being talked about at the moment.’

Forbes Dewey of the Massachusetts Institute of Technology, a close
colleague of Kitney’s, agrees. Dewey cites the case of London’s Hammersmith
Hospital and its much-vaunted Picture Archiving and Communication System.

PACS is a ground-breaking scheme for digitising X-rays so they can be
stored on an optical disc. Radiologists say that a third of X-rays simply
get lost in conventional filing systems. Holding the images electronically
and being able to call them up instantly on a computer terminal should
avoid this problem, and could also help to cut costs by reducing the space
needed for storage, and the time spent retrieving and filing the images.
Hammersmith began installing a Siemens-built archiving system in 1990, and
despite teething problems, is hoping to go live by the end of the year.

Dewey says the Hammersmith system will be fine when it is complete,
but at the moment the system uses Siemens’ own data standards, and needs
dedicated Siemens terminals to display its images. This will make it difficult
for Hammersmith to move images across to specialist graphics workstations
where extra processing could be carried out, or to transmit images to hospitals
which adopt a different brand of PACS hardware. The lack of common standards
has already meant that a significant part of the project’s £13 million
budget has gone on hand-crafting links between the image database and the
hospital’s two existing computerised patient records systems so that the
patient’s name could be attached to each X-ray.

‘It is taking several years just to integrate the text-based hospital
information systems with the image database of PACS. What if in the future
you wanted to integrate ultrasound images to the system? Or to network
it to hospitals using systems other than a Siemens one?’ asks Dewey. ‘It’ll
be expensive if not impossible.’

Concern that a lack of common standards will stop medicine making more
than a first tentative step into the world of computer imaging and electronic
surgery prompted Kitney and Dewey to form the International Consortium
for Medical Imaging Technology (ICMIT), which held its inaugural meeting
in Boston, Massachusetts this week. Kitney and Dewey have recruited 21 other
top researchers to the cause and they are beginning the difficult job of
constructing a standards framework. A hint of what the task ahead involves
can be seen in the framework that ICMIT is proposing.

To get a better idea of the standards needed, Imperial College is building
a ‘technology demonstrator’ – a mock-up of a medical workstation of the
near future. This includes a standard computer running the widely used UNIX
operating system, with a graphical interface designed for medical staff,
and a piece of software known as the Medical Image Display and Analysis
System. MIDAS can accept image data from a range of medical scanning devices,
and then automatically convert this into a format that can be displayed
on the computer screen. The user does not need to be involved in the conversion
project, and as MIDAS is based on current software development standards,
the system will run on different makes of computer system. MIDAS also provides
medical staff with a multimedia environment, capable of dealing with sound
and video as well as text and images.

Mix and match

‘If you think about all the things you will want to be able to display
at a terminal, any standard is going to have to be a multimedia one,’ says
Kitney. ‘As well as X-rays and MR (magnetic resonance) scans, you would
want to be able to show EEG (electroencephalogram) traces. You would want
to be able to include voice messages from the doctor who first examined
the patient. You would want to be able to display the pulmonary sounds recorded
with stethoscopes or moving images from ultrasound and PET (positron emission
tomography) scans. There are so many different modalities that you must
start with a broad framework.’

One example of the type of application that Imperial College is working
towards is an imaging system to aid surgeons repairing arthritic knee joints.
The system not only builds up an animated 3D X-ray of the patient’s joint,
showing the bones and tendons grating against each other as the knee flexes,
but on another area of the computer screen it displays a filmed analysis
of the patient’s gait, allowing the consultant to predict the effect of
corrective surgery.

Kitney says the next step is to start overlaying one kind of image another
rather than having them in separate windows on one screen. For instance,
to diagnose a heart condition, a surgeon might want to take a 3D image of
the heart’s muscle walls and then superimpose an animation of blood being
pumped within the heart: ‘But at the moment you can hardly even get both
images on the same screen let alone match them up. That’s how far there
is to go,’ says Kitney.

ICMIT is hopeful that major computer manufacturers, such as Digital
and Silicon Graphics, will rally to the cause. The consortium is seeking
$15 million in funding to pay for establishing standards over the next five
years. However, one hurdle faced by ICMIT is the different pace of development
in each imaging field – ultrasound, magnetic resonance and computerised
tomography. It makes it difficult to coordinate work on standards within
one imaging discipline, let alone between the different approaches.

The field of ultrasound offers a good example of how rapidly imaging
is advancing. Ultrasound works by measuring the phase change of high frequency
sound pulses reflected through the soft tissues of the body. Because of
the complicated Fourier transform mathematics needed to decode these echoes,
the quality of ultrasound images depends as much upon available computing
power as the sensitivity of the probe used.

Falling computing costs have led to some dramatic improvements in image
quality in just a few years. Five years ago, pregnant mothers would brandish
ultrasound print-outs of what appeared to be aliens looming out of a grainy
gloom. Today, the same pictures are recognisably human. Indeed, on a monitor
the image is clear enough to be able to check the valves of the heart for
defects and even see if the fetus’s bladder is full. Within a few years,
such ultrasound images will be displayed in 3D and vivid colour.

David Evans, a medical physicist at Leicester Royal Infirmary, says
colour-flow mapping – where computers assign colours to different types
of tissue, using red for arteries, blue for veins and white for tumours
– is fast becoming standard on ultrasound equipment. Soon systems will
have the computing power to provide this kind of information on a 3D image.
Existing scanners use only two sensors and so produce a planar, or cross-section,
view. Gaining an extra dimension is simply a matter of adding a third sensor
and performing a further set of calculations. Or better still, ultrasound
could switch to the phased array technology developed for aircraft radars,
where computers ‘steer’ a beam generated by a plate of sensors – making
ultrasound systems easier to use and capable of producing more detailed
images.

Scanning horizons

Not only is the quality of ultrasound images improving by leaps and
bounds, a new generation of thread-like probes is being made which can actually
be inserted into the body. Evans says: ‘With such a probe, you can get right
inside an artery and image the disease lining the walls. Optical methods
are no good for this because, for a start, blood is opaque and you wouldn’t
see much. But ultrasound also gives you a full-depth view – you can image
the tissue that lies beneath the surface. So if you saw a bump in an artery
wall, you could look behind and see if it was a tumour or a nerve structure.’

The uses for such technology are almost unbelievable, says Evans. Unlike
other imaging techniques, such as X-ray or magnetic resonance imaging, ultrasound
needs no expensive shielded rooms and causes no tissue damage: ‘You can
take ultrasound equipment anywhere – all you need is a 13 amp plug. You
don’t need a room full of technicians to run it,’ he adds.

Ultrasound technology is already moving out of its traditional domains
of cardiology and obstetrics. For example, medical staff can use ultrasound
to diagnose cirrhosis from minute changes to the texture of the liver. Such
developments have focused everyone’s attention on ultrasound. But other
medical imaging techniques are undergoing their own revolutions, mainly
through improvements in scanning technologies and the use of graphics workstations
for image enhancement.

One of the most exciting is magnetic resonance imaging. With MRI, patients
are placed inside a powerful magnetic field. This causes the body’s hydrogen
nuclei to emit radio waves, which are used to build an image of tissue density.
As with ultrasound, the quality of MR images is largely dependent on computer
power and so MRI has also benefited from falling hardware prices. Two or
three years ago, scanners took cross sections of a area with a resolution
of 10 millimetres, giving a rather grainy picture. Now the slice size is
3 or 4 millimetres, which gives a much sharper and more accurate 2D image.

New forms of MRI are also beginning to emerge. There is flash MRI which,
as the name suggests, gives an instant snapshot rather than an image blurred
by being captured over time. Then there is MR spectroscopy, which detects
the resonance of nuclei other than hydrogen, MR angiography, which maps
blood vessels, and functional MRI, which captures animated sequences of
movement rather than ‘stills’.

Yet another rapidly evolving image modality is computerised tomography
(CT) – a computer-controlled version of traditional X-rays. Like MRI, CT
produces a sequence of digitised ‘slices’ that can be combined and manipulated
to give coloured, 3D images.

Some idea of the power of CT, once its potential is fully harnessed,
can be gained from the £6 million COVIRA project (Computer Vision
in Radiology), one of several research efforts funded by the European Community’s
Advanced Informatics in Medicine (AIM) programme.

The aim of the three-year project is to help radiologists identify
and target brain tumours using computer-enhanced scan images. A major problem
in treating brain cancer with a radiation beam is that a route has to be
found that avoids sensitive organs like the eyes, brain stem and spinal
cord. What makes planning even harder is that the tumour has to be hit from
at least four different directions so that only the cancerous cells receive
a lethal dose.

Peter Elliott, a physicist at IBM’s UK Scientific Centre in Winchester
and project manager for COVIRA, says the problem with existing CT scans
is that doctors are still working mostly from raw images. ‘What happens
at the moment is consultants have about 30 or 40 slices from a CT scan up
on the screen. Then on each of these slices they have to chase round the
outline of the tumour with a mouse to build up a full 3D picture. This can
take an hour or two. As most radiotherapy departments can only allow half
an hour to plan a treatment, obviously short cuts have to be taken and not
all the at-risk structures might be identified,’ says Elliott.

Full speed ahead

The task of COVIRA is to automate the process which provides an outline
of the tumour. Elliott says this is straightforward using common image
processing techniques such as contour detection. In this technique, computer
algorithms compare adjoining points of an image to find zones of sharp
contrast, which might mark an edge. By using such software to give an approximate
outline, the consultant’s job is simply confirmation, cutting planning time
to 10 or 20 minutes.

Having created a computer image of the tumour, COVIRA’s second objective
is to use the information to control the shape of the treatment beam. ‘Conventional
radiotherapy simply hits the tumour with a rectangular or circular beam
– and because not many tumours are square-shaped, a lot of healthy tissue
gets killed. That’s not a good thing, especially when the tumour is in the
brain. So what we want to do is shape the beam to match the tumour. Of course,
because the beam is going to come from four different angles, that’s four
different shapes that have to be calculated,’ says Elliott.

Being able to shape a beam not only means that cancers will be more
accurately targeted, it will also make it possible to treat the 20 per cent
of brain tumours that are at present untreatable because they lie too close
to a delicate structure or because they have an awkward shape: ‘If you have
a sausage-shaped tumour wrapped round the spinal cord, there’s little hope
of treating it with radiology the way things are,’ says Elliott.

COVIRA not only shows how fast medical imaging is moving, it also demonstrates
how important common standards are to medical technology. The scanners
that capture the initial images cannot manipulate the data to any great
extent. To do anything clever, the images need to be transmitted to general-purpose
graphics workstations which provide cheap computing power. Medical scientists
can now also benefit from the wealth of image processing software originally
developed for entirely different professions, such as oil prospecting or
TV graphics. Yet because scanners use proprietary data formats, moving the
images off the machines can be difficult.

COVIRA illustrates a general truth about the impact of computers and
other emerging technologies on medicine. Mostly, computer power is used
to make existing treatment methods cheaper and safer, rather than to create
entirely novel forms of treatment. As Kitney says: ‘The trend is to do what
we already do – but to do it with better quality and at lower cost.’

Cost is certainly a major concern. In 1990, Lester Thurow of MIT’s Sloan
School of Management estimated that the US spent 12 per cent of its GNP
on health care, much of it on administration. As an example of how costs
could be reduced by using technology, Kitney says it might not be too long
before hospitals routinely transmit images to specialist centres for diagnosis
just as they now send blood and specimen slides for analysis: ‘You could
ship your images around the world in search of the cheapest analysis,’ says
Kitney.

The role of the GP could also change. With the cost of scanning equipment
dropping rapidly, GPs may soon be doing many tests in their own surgeries,
avoiding the lengthy waits for hospital appointments and consultations
with costly specialists. If GPs needed expert help interpreting the images,
they could send them down the telephone line to the appropriate specialist.

But clearly these improvements are unlikely to materialise unless groups
like ICMIT are successful in creating standards to unify the various technologies.
If these are not sorted out soon the medical world risks rediscovering what
the business world learnt from its experience with computers during the
1980s: that building islands of automation brings disappointing returns,
no matter how flashy the individual pieces of technology might appear to
be at the time they are installed. The real power behind the use of computers
lies in the unfettered flow of data and until the issue of standards is
tackled, medical technology will be like a driver trying to accelerate with
the handbrake still jammed on.

John McCrone is a freelance science writer and author specialising in
technology and psychology.

* * *

1: Surgery 2001: cutting links with reality

Richard Satava of the Pentagon’s Advanced Research Projects Agency
lists the key technologies for what he has dubbed Surgery 2001: virtual
reality, microrobots, global communications, 3D imaging and ‘telepresence’
surgery.

Satava says these technologies are developing so quickly that it is
perfectly possible to think about remote surgery on soldiers wounded in
the battlefield or astronauts aboard a space station. His own work with
telepresence systems – where a surgeon wears a dataglove which sends information
to a computer controlling a robot arm, and watches the procedure over a
TV link – began as a collaboration with California research consultancy,
SRI International, which was working on new techniques for surgery. He has
since carried out several ‘proof of concept’ operations – the most recent
being an army field hospital trial conducted this summer in which he operated
on a pig’s intestine laid out on a table in a tent 150 yards away. The operation
simulated gall bladder removal and bowel surgery in a human patient.

Satava says that once such remote surgery techniques have been perfected,
they would find plenty of uses in civilian life. One obvious suggestion
is that virtual reality surgery would allow doctors in big city hospitals
to perform operations on patients in remote rural clinics.

The techniques also opens the way for new kinds of operations. One really
exciting aspect is that it can be scaled up or down. The size of a surgeon’s
fingers can’t be changed but a robot system can be made as large or small
as desired – some work is even being carried out on microrobots that could
be put inside a body.

Keyhole surgery using crude, mechanically controlled instruments is
already having a big impact in medicine. But the development of virtual
reality control – where a surgeon controls the probe with a dataglove, and
receives a probe’s-eye view of the operation on a mounted computer display
– could see even major operations being performed through tiny puncture
holes under a local anaesthetic.

Satava admits there are many problems to be solved before then. Even
with telepresence surgery, where the surgeon watches the movements of the
robot and probe on an ordinary monitor – there is the problem of communications
time lag. With fibre-optic cables, operations over distances of up to 50
miles would pose few problems. But if satellite links were used, this would
introduce a delay of nearly half a second between the time a surgeon made
an incision and witnessed the result. Satava says the robot would need
predictive intelligence so that its movements could be smooth and natural
despite the time lag.

Virtual reality surgery has its problems, too. Satava says the standard
of graphics is now good enough to use in operations. However virtual reality
datagloves still lack the tactile feedback needed by a surgeon.

‘There are at least nine different senses in the fingertips you would
need to model – such as microslip (tiny movements of the finger tips), two
point discrimination, light touch and deep touch, warmth and texture. Gloves
(that simulate these kinds of sensations) are coming along but there is
still some way to go,’ admits Satava.

* * *

2: Other obstacles on the road to electronic medicine

A lack of open data standards is probably the most serious hindrance
to the introduction of new computer-based medical technology – but it is
not the only one. The International Consortium for Medical Imaging Technology
(ICMIT) can name several others.

In the US, there is the spectre of malpractice suits against doctors
and hospitals. MIT’s Forbes Dewey says any procedure which leads to a loss
of original diagnostic data – such as the digitisation of X-rays or the
compression of images for transmission and storage – could lay doctors open
to legal action. But such is the amount of data collected by scans that
retaining all of it would cost a fortune. Dewey says fear of lawsuits is
causing a real hold-up.

Another obstacle to the introduction of technology is the lack of computer
literacy among doctors. Richard Kitney of London’s Imperial College says
this is a particular problem in countries such as Britain, where doctors
tend not to have a hard science background. Peter Elliott of IBM agrees,
saying that specialists tend to feel that the old ways are still the best:
‘It’s a cultural issue. I know many radiologists who instead of manipulating
a CT scan on the screen would prefer to have a print-out so they can go
and hold it up against a light box.’

A third area ICMIT is investigating is economic models. The existing
organisation of Western health services is based on the limitations of old
technology, which required large administrative back-up. New technology
will need new organisations to make best use of it. For instance, GPs may
take on much of the testing now done by hospital consultants, diagnosis
may be electronically farmed out to specialist centres, and consultants
may do a lot of their work over teleconference links. To work effectively
these would all require changes in both the approach and attitude of the
health care professionals, but Dewey says few policy makers are giving the
deeper implications of the new technology much thought.