Paging Dr. Watson
IBM’s Watson Supercomputer Now Being
Used in Healthcare
By Howard Lee
When IBM announced that they were
developing a supercomputing system
called “Watson,” many fans of literary
icon Sherlock Holmes thought of the
London-based detective’s trusted friend
Dr. John Watson—and not the computer’s
actual namesake, IBM founder Thomas
Watson. But the mistake isn’t that far
off after all, for like Dr. John Watson,
IBM’s Watson supercomputer is also
beginning to practice medicine.
Since IBM Watson rose to national
fame, and proved itself by competing and
winning against all time Jeopardy! game
show champions Ken Jennings and Brad
Rutter, the supercomputer has moved on
to practical applications—including
being “taught” to understand the
complexities of healthcare.
Yes, “taught,” and not programmed,
because IBM Watson uses “cognitive
computing,” a completely different type
of computing not found in your desktop
PC. Cognitive computing allows users the
ability to enter mass quantities of
structured and unstructured data from
various sources and ask the computer to
give back a set of structured answers
based only on the most relevant pieces
of the data. While it seems like science
fiction, and a future answer to the
overwhelmingly vast mountains of
currently untapped health data, there
are pilot programs that have recently
launched that use Watson to improve
healthcare processes and treatment.
A Computer with Cognitive Ability
Since the 1940s, computing has relied
on humans programming a set of
instructions into a structured database
and then retrieving the answers from
that data located within the system.
This is done using software to program a
central processing unit (CPU) and using
A to B logic systems. Called the von
Neumann–style, the system of computing
was first laid out by Hungarian-American
mathematician John von Neumann and over
the last 70 years has been used by every
computer company, including IBM, for its
method of creating usable structured
data to perform structured tasks. This
system, while very good at performing a
set of programmed calculations very
fast, can’t interact with its human
counterparts to analyze data, understand
natural languages, or combine structured
and unstructured data into one usable
system. This means that nearly all of
the world’s computer systems are simply
brilliant idiots.
A new computer system was needed that
combined structured data, unstructured
data, natural languages, and data
analysis that could learn from other
systems without the need for a human
programmer to create software for every
scenario. This style of computing system
is cognitive computing, and is the type
being employed by the IBM Watson
cognitive computer system.
When IBM’s computer teams thought
about creating this type of system, the
real task was getting the computer
system to learn from structured and
unstructured data, then combine that
data with natural languages that humans
use every day to come up with answers
that make sense and are completely
useable and practical. It aims to use
the same data used by structured
computer systems, just in more advanced
ways.
While this might sound like
artificial intelligence, cognitive
computing still relies on humans for
part of the work—it is a true
human-machine interface that can create
new computing functions that does not
require tedious software programming for
each new step. Unique to cognitive
computing is the ability for
supercomputers like Watson to learn from
internal and external inputs, and
creates the programming it needs to
solve a given problem. IBM Watson does
this by processing a question in a
similar manner as a human does. It
starts by analyzing the question as
input, then generates a set of features
and hypotheses by looking across data it
has consumed as content. The computer
then seeks the best potential response
to the question.
Using hundreds of reasoning
algorithms embedded within the system,
Watson does a deep comparison of the
language of the question itself as well
as each of the candidate answers. Then
one or more scores are produced for the
algorithms based on the relevance of the
answer, with respect to that algorithm’s
focus area (i.e., temporal, spatial, or
others). It also scores answers based on
contextual relevance. The cognitive
capabilities can then be brought to the
end users through any channel—mobile
device, tablet, desktop computer, etc.
This ability to receive data from a
supercomputer through any device has the
ability to drive positive disruption in
any industry—including healthcare.
For Watson to Thrive,
Providers Need to Connect
Health IT soothsayers believe
that Watson has the potential to
revolutionize healthcare and the
use and management of health
information. But a stark reality
of the present must first be
overcome—how do you get Watson
to talk to different healthcare
organization’s EHRs and access
data in other health IT systems
when hospitals don’t talk to
each other? Watson will only
work if healthcare professionals
are willing to share data with
each other for the benefit of
all and not shutter information
behind locked doors in the name
of protecting proprietary
assets. Health information
exchange must also become more
robust for Watson to succeed.
Health information management
professionals have a role in
facilitating that private,
secure, and authorized
information exchange and should
help establish those data
networks, links, and agreements.
|
Practical Uses for Watson in
Healthcare
The following is an example of how a
cognitive computing system like IBM
Watson could be used to improve
healthcare processes and better analyze
vast amounts of health information. A
doctor gets a visit from a patient who
has diabetes. The doctor determines he
needs to do a blood sugar A1C test, a
blood draw, an EKG, a blood pressure
check, a cholesterol test, and a
physical exam. While this might sound
routine, the way a supercomputer like
IBM Watson analyzes the results is not.
First, the results of a blood sugar
test with a meter are usually logged in
a patient’s diary and not as part of a
database. Since it’s on paper, it is
free text data and thus considered
unstructured data. The A1C is done and
logged into another system to get the
overall three month reading. The blood
draw goes to the lab, where technicians
will look for abnormalities in the blood
that can affect the kidneys, liver,
heart, and cholesterol levels. Blood
pressure is usually done and hand
written in a chart, creating more
unstructured data that is not in the
electronic health record (EHR). EKG
results are checked by a doctor, but
again stored as unstructured data in the
health record. Finally, the physical
exam results are typically written down
by a doctor as a progress note or
dictated exam and not entered as
structured data in the EHR.
Using the typical EHR or other health
IT computer system to give the doctor a
real time diagnosis on this patient in
order to prescribe a treatment would be
very hard given that both structured and
unstructured data has been collected in
a variety of source systems or mediums
and stored in several places. Even for
the human doctor it can be difficult to
determine treatment steps, since the
collected information can get lost or be
misread by the doctor or other
healthcare staffers.
One in five diagnoses are incorrect
or incomplete, and nearly 1.5 million
medication errors are made every year,
according to a 2013 study by Memorial
Sloan-Kettering Cancer Center. The
amount of medical information available
to providers is doubling every five
years, and much of the data is
unstructured, the study says. Healthcare
IT is moving rapidly and developing
other structured and unstructured data
via EHRs and mobile devices like tablets
and smartphones that include data not
entered in any master database. As a
result many doctors can get lost in the
data when trying to treat patients and
determine diagnosis or treatment. Add to
this mix the mountains of white papers
and medical journals that a doctor must
read to stay on top of what is happening
in the healthcare field—plus the ever
growing use of the Internet, blogs,
social media, and healthcare expos to
relay important health information—and
it is evident that there is just too
much information from too many different
sources for any human or typical
computer to analyze in healthcare. There
is too much to do and analyze, and not
enough resources to do it.
The current healthcare system is
doomed to keep making mistakes not
because there is not enough data, but
because there is too much data in too
many places to be useful. Most
healthcare computer systems can only
store and retrieve data, but not do much
more beyond that. Another shortcoming of
a programmed structured data system is
that it can’t understand natural
languages or analyze disparate but
related data in an unstructured form. If
a computer could understand and analyze
both structured and unstructured data
and the relationship between the two, a
doctor would have a system that could
become a true partner in healthcare by
analyzing Big Data and returning the
best and most relevant data for use in
making a diagnosis.
Enter Watson, which through its
cognitive computing has the potential to
look at all of these structured and
unstructured health information sources
and pull together analysis that likely
will improve processes and treatment
plans. While the IBM Watson cognitive
computing system is still very new, it
is not an untested system.
Watson’s Healthcare Case Studies
Below are four case studies that
illustrate how Watson is being used in
fields like cancer research, supply
chain management, and consumer
empowerment to help create better
outcomes in healthcare.
Memorial Sloan-Kettering Cancer
Center
Memorial Sloan-Kettering Cancer
Center (MSKCC), the world’s oldest and
largest private cancer center, is
battling an insidious disease that
strikes one in three women and one in
two men during their lifetimes,
according to data published by MSKCC in
2012. It has become nearly impossible to
find anyone who has not been affected by
cancer in some way.
When trying to find a computer
solution to help analyze their vast
amount of data, MSKCC ran into a
problem. The center has thousands of
cancer patients with different kinds of
cancer, and as many different types of
treatments. With so many treatments and
so much Big Data across as many as 41
different systems, the daunting
challenge of analyzing such disparate
information was one the typical computer
couldn’t handle. Add to that doctors and
researchers creating medical white
papers and journals on the research
being done at MSKCC, and now the
organization had a great deal of Big
Data with no way to really use it to
improve patient outcomes. Looking for a
solution, MSKCC’s CEO Dr. Craig Thompson
joined forces with the IBM Watson team
to teach IBM Watson about their breast
and lung cancer research at the center
and create a system that will allow
MSKCC to use the best available data to
treat their cancer patients. IBM Watson
used its cognitive computing natural
language and decision support system to
find patterns in unstructured
information, mine patient data, analyze
structured data, and look for disease
patterns that most closely approximate
each individual’s case.
Memorial Sloan-Kettering Cancer
Center is now using Watson’s ability to
sort through massive amounts of data,
from clinical knowledge, case histories,
and genomic and molecular data, that
will help oncologists diagnose and treat
an individual’s cancer. But unlike
traditional Big Data computer systems
that simply push data around without
analyzing what data is truly helpful to
the oncologists, Watson actually
understands both structured and
unstructured data and works with a human
counterpart to actually learn from both
Big Data systems and simple doctors
notes. Over time, the hope is that
Watson will become a real part of the
MSKCC oncology team.
University of Texas MD Anderson
Cancer Center
MD Anderson Cancer Center (MD
Anderson) is one of the top cancer
centers in the US. But like MSKCC, they
too have the inability to truly use the
vast research from their oncology team
and combine it with their clinical trial
data to come up with better outcomes for
cancer patients or create targeted
treatment for patients. Within their
healthcare EHR system is all the typical
information that constitutes Big Data,
along with clinical trial data and
mountains of doctors’ private
notes—which make up the backbone of
their research. This has caused a Big
Data divide between MD Anderson’s
doctors and clinical researchers who
work remotely from each other and rarely
share data on the patients they work
with. MD Anderson is now working with
IBM Watson to teach the supercomputer
how to work with its doctors and
researchers. The project is called “MD
Anderson’s Oncology Expert Advisor
(OEA).” The OEA helps doctors and
researchers by integrating the knowledge
from both groups to advance its goal of
treating patients with the most
effective, safe, and evidence-based
standard of care available. Oncology
Expert Advisor provides a 360 degree
view of each cancer patient, which will
help physicians better understand the
patient’s data, history of treatment,
test results, and vital information that
has been hidden in the files of every MD
Anderson medical facility the patient
has visited in the past. By
understanding and analyzing data in a
patient’s profile as well as information
published in medical literature, the OEA
can then work with a doctor to create
evidence-based treatment and management
options that are unique to that patient.
These options include not only standard
approved therapies, but also clinical
trial protocols. MD Anderson’s OEA is
expected to aid doctors to improve the
future care of cancer patients by using
and comparing patients’ data-driven
information—information that was
previously unavailable for complete
electronic analysis.
MD Buyline, Inc.
MD Buyline Inc. has been a leading
provider of healthcare clinical and
technology research for over 30 years
with more than 50 percent of US
hospitals using their solution to track
and improve financial performance across
the healthcare supply chain. A problem
most hospitals face is how can they fill
the procurement needs of their hospital
staff while staying within their budget.
Hospital administrators must find a way
to manage clinical evidence, research,
analysis, and price data from several
different database systems and competing
needs from departments within the
hospital, who are all competing for
resources. This puts a strain on supply
chain management systems, and results in
an estimated $5 billion wasted annually
due to these inefficiencies, according
to an educational paper published by GHX
in October 2012 titled “The Current
State of the Implantable Device Supply
Chain.”
MD Buyline is now working with IBM
Watson to teach the supercomputer how to
understand their supply chain management
and decision support systems. The goal
is for Watson to help deliver a
transformative procurement system that
will enable informative comparison of
medical options. The IBM Watson solution
is expected to drive optimal purchasing
decisions for hospital administrators,
and eventually could offer socially
collaborative and educational support
for all hospital and healthcare teams
worldwide by leveraging a shared base of
information—given that health
information exchange advances to the
point of enabling this connectivity.
Now that healthcare reform has become
part of the landscape, reimbursement
models will be shifting to an
outcomes-based approach. It is more
critical than ever that healthcare
providers improve outcomes. MD Buyline’s
procurement advisor is working to help
deliver these improved outcomes through
its scalable IBM Watson cognitive
computing platform. MD Buyline’s
application is intended to empower users
and aid in finding unbiased clinical
evidence, research, and price
information.
Welltok, Inc.
Welltok, Inc. is a consumer-centric
healthcare company that is changing the
way patients and their caregivers deal
with healthcare by offering programs
that reward consumers for taking charge
of improving their healthcare. Welltok’s
CaféWell, part of their Health
Optimization Platform, is a web-based
community that provides consumers with
healthcare resources, social networking,
gaming, and site personalization. The
site is built to create a new supply
chain, connecting health plans and
health systems to consumers through an
organized ecosystem of health and
wellness resources.
CaféWell’s goal is to help healthcare
managers benefit from increased consumer
engagement, member retention, and
improve brand affinity. Consumers are
rewarded for starting a healthy
lifestyle. Most consumers lead busy
lives that make it hard to follow the
right health decisions. Many consumers
also find it difficult to get the right
information to make proper personalized
healthcare choices. The CaféWell
Concierge, which is powered by IBM
Watson, personalizes the healthcare
experience by using IBM Watson’s natural
language and analysis abilities, as well
as the ability to learn from the
consumer. Watson helps empower the
consumer to make positive healthcare
decisions by providing customized
guidance on activities and behaviors
tailored to a user’s interests and
aligned to their incentives. The
application is also available through
mobile devices, tablets, and personal
computers. Welltok is one of the first
companies to take advantage of Watson
and use it for consumer-facing
applications.
As Watson’s Ecosystem Grows,
Competitors Emerge
IBM just announced the creation of
the IBM Watson Ecosystem that will help
major industries like travel, retail,
and healthcare leverage Watson’s
cognitive computing. Those working with
IBM will get open access to the platform
that will allow them to build customized
applications. IBM business partners will
be able to develop embedded applications
on the Watson Developer Cloud, and have
access to the Watson platform and its
associated tools and methodology. The
hope is IBM’s partners can take Watson
and develop a wide array of products
that leverage its supercomputing
abilities. The ecosystem would bring
Watson to the masses, and potentially be
made available to end clients in
business models such as business to
business, business to consumer, or
consumer to consumer.
Many technology experts expect that
giving developers access to the Watson
Hub—which includes cloud access, a
content store, Watson hosting services,
and tech support—will drive the
development of more healthcare
applications. Soon it is expected that
Watson will be able to work with single
system EHRs or multiple disparate EHRs
and securely access the vital health
information contained therein for
optimizing research. Such access would
need to come with various privacy
measures, which would take time to
develop.
When it comes to cognitive computing,
IBM Watson is the leader at the moment.
But competitors like Microsoft and Apple
have also begun working on their own
systems. The competition is likely to be
a good thing for healthcare, helping
foster innovation in all cognitive
computing system products.
So while it may be some time before
your local hospital pages Dr. Watson for
advice on your medical ailments, these
projects utilizing the technology show
that the use of supercomputers in
healthcare has already begun.
Howard Lee (wireheadtec@gmail.com)
is chief information officer at Wirehead
Technology.
Article citation: Lee,
Howard. "Paging Dr. Watson:
IBM’s Watson Supercomputer Now
Being Used in Healthcare." Journal
of AHIMA 85,
no.5 (May 2014): 44-47. |
|