44381810-Intelligent-Decision-Support-System.doc

Intelligent Decision Support System for
retrieval of patient’s information

Index

 Annotational
tools for

 Abstract
 Introduction
 System
description
 Image retrival

building
patient profiles
 Case retrival
 Evalution
 Conclusion
 Bibiliography

the flexibility and speed of
wireless

ABSTRACT

reduce
workers

in

hospitals continue to record,
access and update important
patient

to

improve the quality and

The majority of
healthcare

computing

information

paper

using
charts.

Disparatepatientdata

the

cost

healthcare.

We

of
are

developing

a

mobile

application

that

allows

doctors to efficiently access
accurate real-time patient
information at the point-ofcare.

(clinicalinformation,
laboratory

results

and

medical imagery) is entered
by different caregivers and
stored

at

different

locationsaround the hospital.
This

is

a

cumbersome,

timeconsuming process that
can result in critical medical
errorssuch

as

documents

being

mislaid

or

prescriptions

system

illegible handwriting.

can

assist

caregivers in automatically
searching through very large
repositories

beingmisinterpreted due to

of

previous

patient cases as increasingly
large hospital databases are

Hospitals
everywhereare moving to
integrate health data sources
usingElectronic

The

Health

Record (EHR) systems as
well as takingadvantage of

making manual searches of
such

information

unfeasible.The
performs
prognosis

system
computational

by

providing

decision support for pre-

screening

of

medical

diagnosis.

evolutionary

phase.

The

medical community has an
obligation to the public to

A presenting patient's

provide the safest, most

symptoms can be input to a

effective

portabledevice

the

possible.This is evermore

quickly

achievable with the use of

retrieve the most similar

portable electronic devices

profiles

and

application

and

can
with

diagnoses

known

from

large

healthcare

new

applications

medical
that can be

databases which can be used

delivered

to

treatments,

networks.

diagnosis, test results and

equipped

other information.

computers now have levels

compare

over

wireless
Caregivers

with

mobile

of interaction at the bedside
not

possible

charts

and

with
can

paper

leverage

accurate real-time patient
information at the point-ofcare to make decisions and
take actions more efficiently.
Electronic Health Record
systems
patient

INTRODUCTI

disparate

information

entered

by

caregivers

to

together

ON

allow

as

often

different
be

stored

encapsulated

patient cases in medical
databases. Using portable
Healthcare technology

is

entering

a

new

devices such as Personal
Digital Assistants (PDA's)

and

Tablet

PC's,

all

integrated EHR system can

caregivers can access vital

be

patientinformation as well as

caregivers

other

including

locations in the hospital

relevant

setting to record and input

medical imagery and online

all important patient data,

medical

including

resources

drugreferences,

encyclopedias

used

wirelessly
at

by

different

clinical

directly at the bedside.

information,

up-to-date

In addition to Electronic

status

laboratory

Health Records, the advent

results,

of new wireless technologies

medical imagery.

reports,

medication

is providing many new and
exciting

opportunities.

and

The type of
functionality

weare

Healthcare providers may

providing in the application

now move freely in hospital

includes

buildings

UserInterface (GUI) where

access

with

to

constant

realtime

vital

a

Graphical

caregivers can input and

patient

record

information.Electronic

information

Health Record systems also

straightforward manner and

play an important role in

multimedia annotation tools

long-term healthcare.

for

We propose an

all

patient
in

medical

imagery

a

to

support Communication and

innovative application that

collaboration

allows doctors to efficiently

different

input, query, update, analyse

annotations and other patient

and

electronic

data can be

including

used to support retrieval of

associated medical imagery

patient case histories for

(e.g. X-Rays) on any mobile

comparisonof diagnoses and

or

treatment

compare

patient

records

desktop

device.

Our

between

caregivers.These

procedures

and

effective

integration

of

patient

profiles

for

image

data

with

other

comparative studies to aid

patient

information

both

with more effective medical

within a database and within

diagnosis

and

an adaptive Graphical User

Once an interaction with a

Interface.

patient

is

treatment.

complete

the

current patient's images are

SYSTEM

central repository and the

DESCRIPTION
The system is based
on a three-tier architecture:
client,

server

database.There

and

are

two

primary client components:
a desktop application that is
used by radiologists and a
mobile component that
used

is

The radiologist employs a
suite of image processing
provided

desktop

by

the

application

to

annotate any images they
have

just

relevant

acquired

with

notes

and

information regarding the
patient's

condition.

display previously annotated
images

and/or

annotations

can

their
be

retrieved/updated at a later
date either
by the radiologist or another
doctor

examining

the

patient. Physicians can use
the mobile application on a

from

Digital Assistant (PDA) or a
Tablet PC to retrieve and
131 view patient profiles
and to quickly

a

knowledge base of previous

enter

information about
patient

progress

into

electronic charts in realtime.
The PDA allows

The

system can be queried to
patient

images

Personal

by physicians.

tools

added to their profile in a

physicians to view textual
abstractions

of

current

patient data as well as other
patient case histories and

allows

for

improved

mobile devices may also be

mobility due to its reduced

used to query the central

size and weight. The Tablet

repository of patient profiles

PC

more

with information specific to

user

aparticular illness to retrieve

additional

similar patient case histories

provides

a

comprehensive
interface

and

functionality such as the

that

may

help

ability to view and annotate

diagnosis

medical imagery due to the

comparative assistance.

by

with

a

providing

larger screen size and higher
processing power.All of the
information for each patient

IMAGE
RETRIEVAL

is stored as an integrated
patient profile in a database.
All interactions with the
patient profile by healthcare
workers are recorded by the
system and the profile is
updated accordingly. Both
the PDA and the Tablet PC
can be used wirelessly at
different locations around
the hospital or in other
situations

where

such

facilities
are not normally available
(e.g. in an ambulance).They
may also be used to access
other

resources

such

as

online drug references or
medical

encyclopedias.The

The majority of
current

medical

image

retrieval techniques retrieve
images

by

similarity

of

appearance,using low-level
features such as shapes or by
natural

language

biggest

problem

.

The

arising

from these techniques is the
so-called semantic gap -the
mismatch

between

the

capabilities of the retrieval
system and user needs. In
order to bridge this gap our
application aims to unite
information

about

underlying visual data with
more

high-level

concepts.

For example, capturing a

measure

of

expertise

and

involved

in

the

human

techniques.Using

proficiency

application,

making

can

a

our

the caregiver

retrieve

relevant

diagnosis from an X-Ray us

patient imagery by entering

to relevant image allows

relevant patient data and by

information was selected. .

clicking

on

the

"Search

Patient Images" button on
the interface depicted in
Figure1 . The user can adjust
the weights of the relevant
search fields by using the
slider bars to indicate which
of the parameters are most
relevant. For example a
radiologist may be viewing
an X-Ray image and may
be

having

diagnosing

difficulty
the

in

problem

from the particular image.
Fig. 1. Search for Patient
Information

and

medical

imagery

The X-Ray, however may
remind him of a similar
image he viewed previously

(e.g. the highlighting a

and he may remember some

particular body organ)and

of the details of the previous

also how it was employed in

patient.In this scenario the

the context of that specific

radiologist could input the

diagnosis (e.g. inferred from

details of the

an added annotation).

previous patient as search

The approach

parameters

to

the

allows us to capture and

application.The application

reuse

will then filter out this

best

practice

patient's profile as well as

image

in

any similar profiles from the

appropriatefashion

EHRS. The radiologist can

diagnosing

the

an
while
patient

.

then compare the current
image to images from these
previous case histories to try
to find any similarities. If
any of the similar images
had been annotated while
those patients
were being diagnosed the
radiologist may study these
notes for extra information
regarding the specific injury
or illness.

ANNOTATION
TOOLS FOR
BUILDING
PATIENT
PROFILES
The annotation tools
provided by the application
are used by radiologists to
annotate medical imagery
with relevant
information.
radiologist

diagnostic
When

a

takes a patient

X-Ray or scan the
can be displayed

image
in the

screen depicted in Figure 2.
Using the tools provided the
radiologist can annotate the

By providing an
interface such as this one we
have achieved the goal of
capturing

important

contextual patient/diagnostic
information

by

situating

intelligent

support

for

gathering it inside a flexible

environment.

This

information

is

implicitly

to

radiographer
burden

collected

annotations

can

previewed before

be
being

shield

the

incorporated as part of the

from

the

knowledge base, and once

explicit

recorded they can be saved

of

knowledge

and uploaded to the image

engineering.From
perspective,

their

as

a

knowledge

parcel

the

image

associated with the patient in

tools

support

question. The system also

them in carrying out their

supports annotation by cut,

task

copy and paste between a

interaction
(e.g.,

a

radiologist

producing a report on the

given

current patient) by making it

images in the dataset, as

easier for them to select

well as any application that

and

supports

highlight

relevant

features, to store insights

image

for

direct

manipulation,
filters,

clipboard
Once the radiologist

has finished interacting with

We have developed
tools

other

functionality.

and to summarize aspects of
their work progress.

and

the medical imagery their

image

entire work process is stored

including

along with all the other

transformation,

patient

data

as

an

highlighting, sketching and

encapsulated patient profile

post-it type tools. The user

in the knowledge base.

can add media annotations
to images as a whole or to
particular highlighted image
aspects.

Currently,

CASE
RETRIEVAL
As the system builds

the

system supports annotation
by text audio and video. All
textual, audio and video

up

encapsulated

user

interactions,another type of
retrieval

is

enabled,

retrieving

entire

previous

entire case history in a

patient case histories. This

manner

enables

enough to allow multiple

a

physician

radiologist

to

look

or
for

previous patient analyses .

results

that
to

is

compact

be

simultaneously..

viewed
Figure

3

shows
an example of our results for
retrieved case histories.Each
row represents a patient case
history

It

includes

the

matching percentage score
between
the current query and the
similar patient profile. The
user can click on the "Open
Profile" button to view the
The caregiver can
retrieve relevant patient case
histories by entering the

full case history (including
any medical imagery) of that
patient.

relevant patient details to the

EVALUATI

interface and by clicking on
the

"Search

for

Patient

ON

Profiles" button. As in image

In an initial

retrieval the user may adjust

evaluation

the relevance of the search

conducted testing with a

fields

dataset of 100 encapsulated

by

moving

the

associated slider bars.
challenge in
previous

case

we

have

patient profiles . In this
One

preliminary evaluation we

retrieving

conducted experiments to

histories

test the previous patient case

has been how to present an

retrieval capabilities of the

or "not relevant" to the

application. In this approach

search query. These ratings

we

in

were then compared to the

showing that the system is

clusters we had outlined

capable of capturing and

earlier to examine if the

deciphering expert medical

results were appearing in the

knowledge and we were

relevant categories.

were

aiming

interested

to

show

that

recommendations made by
the system based on similar
patient data. We were also
interested in demonstrating
the ability of the application
to

facilitate

effective

knowledge and data sharing.
The 6
selected cases were then
input as search parameters
to

application.The

information,

symptoms,

diagnosis and treatments for
each case was entered to the
system and similar cases
were retrieved by pressing
the "Find Similar Cases"
button. The cases retrieved
by

the

displayed

application
in

the

and
results

screen were then analysed.
Each

returned

case

was

marked as either "relevant"

This process was repeated
for each of the 6 cases .In
order to graph the results for
our

prior

algorithms
precision

case
we
and

retrieval
employed
recall

metrics.
Figure 4 shows the average
precision and recall values
for the results of the 6
queries. From the graph we
observe that
the system is performing
case retrieval accurately. we

also

observe

a

linear

increase in recall as the

prognoses

with

greater

speed and efficiency.

number of results returned

It facilitates

increases indicating that all

knowledge

relavent cases are being

supports

retrieved for each query.

communication

about the

most

ways

CONCL

current

system

hospitals

where

in
by

doctors enter
papercharts
consuming and
not

information,

facilitate
including

imagery,

are

different

locations

stored

in
and

valuable time is often lost
trying to correlate data in
order to diagnose and treat
patients.This

system

such

issues

doctors

can
by
with

instant access to information
that will allow them to make
critical

histories using case-based
reasoning

techniques.It

adds more value to imagery
combining it with patient
thorough

communication,

examination and diagnosis.

knowledge sharing.types of

providing

in

records to support more

cumbersome, time-

address

effective

and image transmission by

information using

does

effective

linking similar patient case
The

is

and

which to treat patients by

USION

patient

sharing

decisions

and

Our initial evaluation of the
system has produced very
promising
system
capture

results.
can

The

effectively

important

information

and

sucessfuly

used

patient
can

be

retrieve

similar previous patient case
histories

that

can

offer

support

to

useful realtime
decision

physicians at any location in
the

hospital

setting.

We

intend to conduct trials with

domain experts in the near
future.

We

intend

to

incorportate a facility to
record relevance feedback
from physicians to improve
the

usability

of

the

application for the expert
users.

BIBILIOGRAPHY
1:www.medicalimagecomputin
g.com
2:www.medical-imageprocessing.info