3. Course Topics
• Human and Clinical Activities
• Medical Informatics – Definition and Concepts
• Computer Systems
• Signal Processing
• Patient Data and Input Strategies
• Patient Records
• The Internet
• Communication Systems
• Information Retrieval Systems
• Computed Radiology, Medical Imaging and Image Processing
• Hospital Information Systems
• Decision-making Strategies
• Statistical Methods
4. Further Reading
• Shortliffe EH, Perreault LE, Wiederhold G &
Fagan LM : “Medical Informatics – Computer
Applications in Health Care”. Addison-Wesley
Publ., 1990.
• Van Bemmel JH, Musen MA: “ Handbook of
Medical Informatics”. Mieur Publ., 1999.
– http://www.mieur.nl/mihandbook/r_3_3/toc/toc.htm
• Jenders R, Sideli R, Hripcsak G: “MDIF G4001
Introduction to Medical Informatics - Online
Lecture Notes”.
– http://www.dbmi.columbia.edu/~hripcsa/textbook
5. AIMS OF THE COURSE
• Provision of an introductory view to computer
science for medical professionals
• Basic coverage of:
– Windows
– Office
– Programming languages
– Internet with specific emphasis on technology and
various applications in medicine
6. AIMS OF THE COURSE
• Discussion of important topics dealing with the
use of computers in medicine for:
– Research
– Collaboration
– Communication
• Addressing the use of computers as tools for:
– Clinical work
– Lifelong learning and education
11. MEDICAL INFORMATICS
The term medical informatics dates from the second half of the 1970s
and was borrowed from the French expression informatique médicale.
12. MEDICAL INFORMATICS
Before that time, other names were used (and are sometimes still in use),
such as
medical computer science,
medical information science,
computers in medicine,
health informatics,
and more specialized terms such as
nursing informatics,
dental informatics.
13. MEDICAL INFORMATICS
These terms have their parallels with similar ones in areas outside health
care, such as
computer science,
information processing,
informatics,
and in specialized areas, for instance:
computational physics,
computational linguistics,
artificial intelligence.
14. MEDICAL INFORMATICS
In informatics as such, one may discern three different layers of research:
fundamental computer science,
applications-oriented informatics,
applied informatics.
Research for the realization of medical information systems mainly
belongs to the third category.
15. MEDICAL INFORMATICS
The more usable and more comprehensive the software tools that come
out of research in computer science, the better medical informatics can
direct its efforts on applications-oriented projects where specific
knowledge is required in the specialized area.
16. TWO DEFINITIONS OF MEDICAL INFORMATICS
1. Medical information science is the science of using system-analytic
tools . . . to develop procedures (algorithms) for management, process
control, decision making and scientific analysis of medical knowledge.
2. Medical Informatics comprises the theoretical and practical aspects of
information processing and communication, based on knowledge and
experience derived from processes in medicine and health care.
17. Why Study Medical Informatics
• Well-trained health care professionals raise the
quality of information processing.
• The quality of information processing influences
the quality of health care itself.
• Therefore, for a systematic processing of
information in health care, health care
professionals who are well-trained in medical or
health informatics are needed.
18. Why Study Medical Informatics
• Medical or health informatics education has
become an integral part of education and training
for physicians, nurses, and administrators in
different countries all over the world.
• In addition, medical informatics courses are
offered to informatics students, and dedicated
programs are organized for specialists in medical
informatics.
26. Components of a Computer System
• Hardware (the physical equipment) including:
– CPU
– Data storage devices
– Terminals
– Printers
• Software
– This embraces the computer programs that direct hardware
to process input data and stored information.
• Users
– These are the people who interact with the software and
hardware of the system.
36. Functions of a Computer System
• Data Acquisition
• Record Keeping
• Communication and Integration
• Surveillance
• Information Storage and Retrieval
• Data Analysis
• Decision Support
• Education
41. Definitions
• Bandwidth:
– It is the range of significant frequencies contained in
a signal or which a system can deal with.
• Quantization:
– It is the number of bits represented digitally by one
sample
42. Examples of three biological signals (ECG, top; spirogram, middle; and EEG,
bottom) with their frequency spectra given at the right. Of the frequency spectra,
only the signal power per frequency component is shown.
43. A vectorcardiogram together with its projections on the three orthogonal planes:
the horizontal plane (XZ), the frontal plane (XY), and the sagittal plane (YZ).
44. Frequency spectra of an ECG plus disturbance and of the QRS complex, the P-
wave, and the disturbance separately.
46. Report of a computer-processed ECG as generated by a commercial system that
interprets ECGs by the modular ECG analysis system MEANS (display from
Cardio Perfect system).
47. Evoked potentials in seven EEG leads resulting from stimulation by light flashes
with a frequency of 1.9 flashes per second. The potentials are the result of the
summation of 128 evoked potentials and the interval before the light flash is 10
msec.
49. Data vs. Knowledge
• Data = observation (heart murmur: description
or tape recording)
• Knowledge = interpretation of data
– the interpretation itself (defective heart valve)
– methods to interpret data (process of diagnosis)
• Knowledge at a certain level (A) may become
data at a higher level (B)
• Information includes both knowledge and data
50. Importance of Data
• To characterize patients
• To enable diagnostic and therapeutic decisions
51. Example Questions that Data Can
Answer
• history?
• symptoms?
• examination findings?
• changes in symptoms and signs over time?
• changes in physiologic function over time?
• previous treatment?
• rationale for treatment?
52. Problems with Data
• Costly
• Risky
• Uncertain
– inaccuracy of test
– gaps in scientific knowledge
– poor memory
53.
54.
55. CODED DATA
• It is an identifying information (such as name or
social security number) that would enable the
investigator to readily ascertain the identity of
the individual to whom the private information or
specimens pertain has been replaced with a
number, letter, symbol, or a combination thereof
(i.e., the code)
• A key to decipher the code must exists to
enable linkage of the identifying information to
the private information or specimens.
56. CODED DATA
• can be viewed as a tuple (in form of a relation or a fact
with several components)
– patient identifier (e.g., medical record number=1234)
– name of parameter (e.g., potassium)
– value (e.g., 4.5 mg/dl)
– time (e.g., 1993-01-04 14:35:23)
– modifiers (e.g., likely)
• like narrative, can be based on paper or computer
– have been using forms for years (e.g., name and age)
– structured entry form and entry clerk
– direct entry
– automatically or manually code narrative text
57. NARRATIVE (Natural Language) DATA
• Written in natural language
• Bulk of data is in this form
– Admission notes
– Progress notes
– Discharge summary
– Nursing notes
– Radiology reports
– Pathology reports
• English plus technical terms specific to the field
• Idioms/phrases
• Shorthand narration
– "dyspnea on exertion" (DOE)
• Can be entered on paper or computer
– write in chart
– write on form and transcribe
– dictate and transcribe
– word processor
– "clickable" templates
58. ADVANTAGES OF NARRATIVE VS.
CODED DATA
(independent of use of computer)
• Narrative
– universal and familiar
– little training, except for overall format of note
– expressive, even for new and unexpected situations
• Coded
– encourage complete entry (phone number field)
– improve ability to locate and read data
– improve data classification
– force the use standard vocabulary (ICD-9, DRG)
– aids retrieval, research, billing
59. 31.02 Data Entry Communication Styles
Input Style Description
Form filling When the same type of data have to be entered repeatedly, the
use of forms (having the appearance of a paper form) can
facilitate data entry
Question and answer
dialogues
Here the system prompts with a question to which users
respond (e.g., by selecting from preset choices). This is a style
for data entry which is useful for novices but often frustrating
for expert users.
Menus Menus provide users with a limited set of options from which
to choose. Types of menus include:
fixed menus (which stay in place),
pull down menus (which are dragged from a title or menu
bar), and
pop-up menus (which appear when a user clicks a particular
part of the screen)
60. 31.02 Data Entry Communication Styles
Input Style Description
Natural language This is to allow users to enter data in unrestricted natural
language. This has proven to be a difficult problem to solve
because of the complexity of natural language. A solution is to
limit the domain to a subset of language.
Direct manipulation Instead of making users remember commands, direct
manipulation allows the users to select objects and actions by
using a pointing device
61.
62.
63. 12.01 Type of Data Used for Patient Monitoring in Different ICUs
Continuous Variables Sampled Variables Coded Data Narrative
text
Temperature
Central
Peripheral
Patient observations
Color
Pain
Position
etc.
All other
observations
or
interventions
that cannot
be measured
or coded
Cardiac
ECG
Heart rate (HR)
HR variability
PVCs
Blood pressure
Arterial/venous
Pulmonary
Left/right
atrial/ventricular
Systolic/diastolic
Per beat/average
Systolic time intervals
Blood chemistry <UL
Hb
pH
PO2
PCO2
etc.
Interventions
Infusions,
Drugs
Defibrillation
Artificial
ventilation
Anesthesia
etc.
64. 12.01 Type of Data Used for Patient Monitoring in Different ICUs
Continuous Variables Sampled Variables Coded Data Narrative
text
Respiratory
Frequency
Depth/volume/flow
Pressure/resistance
Respiratory gases
Fluid balance
Infusions
Blood plasma
Urine loss
Neurological
EEG
Frequency
components
Amplitudes
Coherence
65. 31.01 Input Devices
Device Description
Keyboard Most commonly used, especially for data entry
Mouse Very useful device to point to the display screen
but with limitations for use in health care
Trackball Is a rotating ball and similar to the mouse to point
to the display screen
Graphics tablet Is a touch-sensitive surface which lies apart from
the display screen. Can be operated by using
finger, pencil, or stylus. Is sometimes used for
data entry in computer-based patient records
66. 31.01 Input Devices
Device Description
Touch screen Does not require any other device but rather the
user touching the display screen with his or her
finger
Pen A lightpen allows the user to point at spots on the
display screen. More advanced versions of
computer pens allow users to handwrite on the
screen
Speech Voice input to the computer, allowing to use
natural language. In the future this holds much
promise for health care.
67. Example of a screen of the Pen&Pad system that helps the user to select the
symptom or complaint to be described. The list on the left presents complaints of
a general nature. The symptom list on the right has a specific focus, depending on
a location selected by the user, in this case the chest.
69. Purpose of the Medical Record
1. Organization/integration of data to reveal:
relation among problems
global picture of patient by avoiding missing details
2.Legal
A legal record for litigation and accreditation
3. Communication
This is important as "health care team" expands to include many professions (PT,
OT, RT, VNS), thus providing continuity of care
4. History
A historical record for clinical care
5.Quality assurance
6.Financial
coding for bills
proof of charge validity
7.Prospective research
screening
tracking patients
8.Retrospective research
looking across existing records
70. Sources of Data in a Medical
record
• Clerks
• Primary physicians
• Consultants
• Nurses
• Therapists
• lab reports
• Machines
71. Fundamental Issues Concerning Patient
Records
• Data capture
– getting data from external sources can be difficult
• Data input
– entry can be time-consuming and expensive "free text" vs.
structured entry
• Error
– can occur at any point in the entry model
• Completeness
– proportion of observations actually recorded (30-100%,
depending on type of observation)
• Correctness
– proportion of recorded observations that are accurate (67-
100%)
• Security
– tradeoff of access vs. security
72. Advantages of Paper Records
• Can be easily carried around
• Much freedom in reporting style
• Easy data browsing
• Require no special training
• Never down as sometimes computers are
73. Weakness of Paper Records
• Finding the record is a problem, because it is
lost
being used elsewhere
• Finding data within the record is also a problem, because the
data is
poorly organized
missing
• Difficulty with reading data (legibility)
• Difficulty with updating data (where to place a missing chart
in the updated record)
• Support of only one view
• Lack of redundancy (need to re-enter data in multiple forms)
• Lack of research facility (cannot search across patients)
• Passivity (no automated decision support possible)
75. Basic Requirements of the
Computerized Record
• wide scope of data
• sufficient duration of use
• understandable representation of data
• sufficient access (number of terminals)
• structured data
77. Advantages of Computer Records
(especially coded)
• Access
– Speed
– Remote location
– Simultaneous use even if just a digitized picture of the paper chart
• Legibility
• Reduced data entry (reuse of data) by eliminating the
need for
– redundant entry
– redundant test ordering
• better organization by imposing structure
• Multiple views, including aggregation
– Summary report
– Structured flow sheet
– Context- dependent display
78. Advantages of Computer Records
(especially coded)
• Automated checks on data entry
– more complete prompting for data
– Range checks (potassium level = 50)
– pattern checks (7 digit phone number)
– computed checks (sum of white counts = 100)
– consistency checks (pregnant man)
– incremental checks (weight increase of 100 lbs)
– spelling checks
• Automated decision support
– Reminders
– Alerts
– Calculations
– Advice on ordering
79. Advantages of Computer Records
(especially coded)
• Cross-patient analysis
– Research
– Stratification of patient prognosis
– Treatment by risks
• Time saving
– Finding data 4x faster in flow sheets vs. traditional
records
– 10% of subjects could not even find certain data
• Data review
– minimizing overlooking uncommon but important
events
80. Disadvantages of Computer
Records
• Initial cost (paper is already paid for)
• Training
• Delay between investment and benefit
• Security concerns
• Computer failures
• Difficulty of data entry
• Requires great coordination of disparate groups
• Cannot always force data into computer's structure
• Data diversity
– accommodating different data elements
– Vocabulary
– Format
– Units
82. Shortness of breath, cough, and fever. Very dark feces.
Exam: RR 150/90, pulse 95/min, Temp: 39.3 oC. Rhonchi, abdomen not
tender. Present medication 64 mg Aspirin per day. Probably acute
bronchitis, possibly complicated with cardiac decompensation.
Bleeding possibly due to Aspirin.
ESR 25 mm, Hb 7.8, occult blood feces +.
Chest X-ray: no atelectasis, slight sign of cardiac decompensation.
Medication: Amoxicillin caps 500 mg twice daily, Aspirin reduce to 32
mg per day.
Feb 21,
1996:
No more cough, slight shortness of breath, normal feces. Exam: slight
rhonchi, RR 160/95, pulse 82/min. Keep Aspirin at 32 mg per day.
Hb 8.2, occult blood feces.
Mar 4,
1996:
Time-Oriented Medical Record
83. Visits
Feb 21, 1996: Shortness of breath, cough, and fever. Very dark feces.
Exam: RR 150/90, pulse 95/min, Temp: 39.3 oC.
Rhonchi, abdomen not tender.
Present medication 64 mg Aspirin per day. Probably acute bronchitis,
possibly complicated with cardiac decompensation.
Bleeding possibly due to Aspirin.
Medication: Amoxicillin caps. 500 mg twice daily, Aspirin reduce to 32 mg per day.
Mar 4, 1996: No more cough, slight shortness of breath, normal feces.
Exam: slight rhonchi, RR 160/95, pulse 82/min.
Medication: keep Aspirin at 32 mg per day.
Source-Oriented Medical Record
84. Source-Oriented Medical Record
Laboratory
tests
ESR 25 mm, Hb 7.8, occult blood feces +.Feb 21, 1996:
Hb 8.2, occult blood feces.Mar 4, 1996:
X-rays
Chest X-ray: no atelectasis, slight sign of cardiac
decompensation.
Feb 21, 1996:
85. Problem-Oriented Medical Record
Problem 1: Acute bronchitis
Shortness of breath, cough, and fever.S:
Feb 21,
1996
Pulse 95/min, Temp: 39.3o.CO
Rhonchi. ESR 25 mm.
Chest X-ray: no atelectasis, slight sign of cardiac
decompensation.
Acute bronchitis.A:
Amoxicillin caps. 500 mg twice daily.P:
No more cough, slight shortness of breath.S:Mar 4, 1996
Pulse 82/min. Slight rhonchi.O
Sign of bronchitis minimal.A:
86. Problem-Oriented Medical Record
Problem 2: Shortness of breath
Shortness of breath.S:
Feb 21,
1996
Rhonchi, RR: 150/90.O
Chest X-ray: no atelectasis, slight sign of cardiac
decompensation.
Minor sign of decompensation.A:
Slight shortness of breath.S:Mar 4, 1996
RR: 160/95, pulse 82/min.O
No decompensation.A:
87. Problem-Oriented Medical Record
Problem 3: Dark feces
Dark feces.S
Feb 21,
1996
Present medication Aspirin 64 mg per day.
Abdomen not tender, no blood on the glove at rectal
examination Hb 7.8.
O
Intestinal bleeding possibly due to Aspirin.A
Reduce Aspirin to 32 mg per day.P
Normal feces.SMar 4, 1996
Occult blood feces.O
No more sign of intestinal bleeding.A
Keep Aspirin at 32 mg per day.P
89. What is the Internet
• The Internet is a worldwide collection of networks that
support personal and commercial communications
and information exchange.
• It consists of hundred thousands of separate networks
passing data to each other using a protocol called
TCP/IP.
• The networks are interconnected and are accessed
daily by millions of people.
• Each computer on the Internet has a unique identifier
referred to as the IP Address.
90. How to connect to the Internet?
Phone line:
connecting to the Internet from home using modem to
connect to the Internet by using the telephone line whether
dial-up, or DSL or ISDN.
Leased Line:
another way to connect to the Internet is by connecting to
Local Area Network (LAN) by using a network card (or
Ethernet card).
91. Components of the Internet
• Personal computer
• Web browser software to access the web
• A connection to an Internet Service Provider
(ISP)
• Servers to host the data
92.
93. Services of the Internet
• Email
• WWW
• FTP
• Telnet
• Newsgroups
94. World Wide Web (WWW)
• The World Wide Web (WWW), or Web is an
information retrieval system based on
technology that organizes information into
pages.
• It is thus an online document system that
supports links between documents.
• The Web is only a subset of the Internet.
95. World Wide Web (WWW)
• There are several client applications called
browsers (e.g., Netscape, Internet Explorer,
Mosaic), which
– Facilitate network connections to WWW services
– Display the documents provided by a WWW server
• A browser is a software tool that resides on the
computer enabling the user to view WWW
documents and access the Internet taking
advantage of text formatting, hypertext links,
images, sounds, motion, and other features.
96. World Wide Web (WWW)
• When a WWW document is displayed to a
client, all words in the document that provide
links to other documents are highlighted or
underlined.
• Selected such a highlighted or underlined
“string” results in a message to a WWW server
indicating that the linked document should be
retrieved and sent to the client.
97. World Wide Web (WWW)
• WWW documents may consist of
– Text
– Images
– Video
– Sound
• For all these types of information, the browser
must provide means for playing or displaying
the information.
•
98. WWW Servers
• An attractive feature of WWW documents is
that a phrase can be linked to a document on
another server.
• For example, a WWW document on a server in
New York can have a link to a document in
Amsterdam.
• The user of WWW is not aware of the physical
location of the servers in the Internet.
99. WWW Servers
• An example of the WWW server is the Virtual
Hospital.
• This WWW server provides a large number of
documents and a large number of hypertext
links that are relevant to healthcare.
100. Web Pages
• Web pages are connected by
hyperlinks/hypertext – indicated onscreen by
boldfaced and underlined text or by selectable
icons.
• Clicking on a hyperlink will retrieve another
document.
101. Web Sites
• A Web site consists of one or more Web pages
that relate to a topic or business.
• The first page a user sees in a Web site is
called the home page.
• The Web sites are hosted by Web servers.
102. Hypertext
• Hypertext allows a text area, image, or other
object to become a "link" (as if in a chain) that
retrieves another computer file (another Web
page, image, sound file, or other document) on
the Internet.
103. URL
• The Uniform resource locator (URL) is a unique
identifier representing the location of a specific Web
page on the Internet.
• A hypertext link in a Web page contains one of the
URLs.
• Clicking on that link will invoke a request to retrieve a
unique document identified by that URL.
• Example:
http://www.guc.edu.eg/faculties/index.html
104. URL
• http (Hypertext Transfer Protocol) is a set of rules
transferring files (text, graphic images, sound, video,
and other multimedia files) on the WWW.
• www This indicates a page on the World Wide Web.
(Sometimes "www“ is missing.)
• guc.edu.eg is the domain name of the URL.
• /faculties/index.html is the path to the file.
• The document index.html is stored in the folder
faculties on the Web Server.
105. IP-Address (Internet Protocol
Number or Address)
• A unique number consisting of 4 parts
separated by dots, e.g. 165.113.245.2 . Every
machine that is on the Internet has a unique IP
address.
• The former part of the IP address (165.113) is a
unique identifier of the network, while the latter
part identifies a computer in the network.
• Most machines also have one or more domain
names that are easier for people to remember.
106. Domain Name
• The domain Name is an ordered group of
symbols, separated by periods.
• It identifies an Internet server.
• The domain name www.guc.edu.eg points to
all computers in the GUC.
• Examples of some Domain names are:
www.yahoo.com
www.informatik.uni-ulm.de
Back
107. Top-Level-Domain
• The last part of the domain name (.com, .edu, .eg, .de)
is the most general part of the domain name and is
called the Top-Level-Domain (TLD).
• The TLD usually reflects the purpose of an
organization or entity.
• Examples of TLDs are:
edu: for educational institutions
com: for commercial organizations
gov: for governmental institutions
• The TLD can also indicate a country code. For
example, eg refers to institutions in Egypt, de refers to
institutions in Germany.
109. Networks
• What the need for so many computers (lab, radiology,
... etc.), thus forming a network
– Many small computers can do more than one big computer
(even considering the cost of connecting them)
– Local autonomy provided
– Less vulnerability to failure
• Network goals
– Communication between users
– Sharing resources and reducing costs – e.g. printer
– Reliability - through hardware redundancy
– Sharing of data
– Centralization of administration and support
110. More specifically, computers that are part of a
network can share:
• Documents (memos, spreadsheets, invoices, and so on).
• E-mail messages.
• Word-processing software.
• Project-tracking software.
• Illustrations, photographs, videos, and audio files.
• Live audio and video broadcasts.
• Printers.
• Fax machines.
• Modems.
• CD-ROM drives and other removable drives, such as Zip and
Jaz drives.
• Hard drives.
113. Networks
• Definitions:
– A network is a connection of 2 or more devices capable
of independent control
– baud rate = switching-values/second (bits/second, not
bytes/sec)
– LANs are kilometers wide, tightly coupled
– WANs cover unlimited distance, but slower
• Connections:
– Without a LAN: N(N-1) connections required. Usually
one cannot reuse connections.
– With a LAN: N connections are required.
114. Networks
• Medium: physical link
– Twisted pair having a bandwidth of 4 MB. They are
cheapest, like telephone wire
– coaxial cable (bandwidth =100 MB)
– fiber optics (bandwidth =1 GB). They are fastest,
producing no radio interference
• Interfacing with the computer
– network interface (adapter) card (NIU) connects
wire to computer
115. Local-area Networks (LANs) and
Wide-area Networks (WANs)
• Network communication within an organization is
referred to as a LAN. LAN Characteristics:
– Physical distance between computers is restricted
– Implements one network protocol
– Has a high communication bandwidth
• WAN Characteristics:
– Span much larger areas, even worldwide
– Provide less bandwidth than a LAN
– Several LANs can be interconnected (by means of so-called
gateways) to allow for communication far apart (thus forming
a WAN)
119. Bus Network Topology
This is a network architecture in which a set of
clients are connected via a shared
communications line, called a bus.
120. Bus Topology
• Advantages
– Easy to implement and extend
– Requires less cable length than a star topology
– Well suited for temporary or small networks not requiring
high speeds(quick setup)
– Cheaper than other topologies
• Disadvantages
– Limited cable length and number of stations
– If there is a problem with the cable, the entire network goes
down
– Maintenance costs may be higher in the long run
– Performance degrades as additional computers are added
or on heavy traffic
– If many computers are attached, the amount of data flowing
causes the network to slow down
– It works best with limited number of nodes
122. Star Network Topology
• They are one of the most common computer
network topologies.
• In its simplest form, a star network consists of
one central switch, hub or computer which
acts as a conduit to transmit messages.
123. Star Topology
• Advantages
– Good performance
– Easy to set up and to expand
– Any non-centralized failure will have very little effect on
the network
– Easy to detect faults
– Data Packets are sent quickly as they do not have to
travel through any unnecessary nodes
– Used for centralized control
• Disadvantages
– Expensive to install
– Extra hardware required
– If the hub/switch fails the entire system is affected
125. Network Protocols
• Before two computers can exchange
information via a network, they should agree
about the format of the information to be
exchanged.
• Nowadays, there are several standard protocols
for that purpose.
• The best known of these protocols is the
Transfer Control Protocol/ Internet Protocol
(TCP/ IP).
126. Network Protocols
• Solutions have been provided for worldwide
communication between computers that use
different network protocols.
• These are in the form of network gateways.
• Gateways translate one network protocol into
another.
128. • A bibliographic or library database is a
database of bibliographic information. It
may be a database containing information
about books and other materials held in a
library (e.g. an online library catalog, or
OPAC) or, as the term is more often used,
an electronic index to journal or magazine
articles, containing citations, abstracts--
and often either the full text of the articles
indexed, or links to the full text.
129. Purpose of Information Retrieval Systems
• The purpose of information retrieval is to provide
information that changes the knowledge state of a
user so that this user is
– better able to perform a present task (solve a problem, make
a decision),
– is better prepared to perform future tasks,
– is better able to assimilate information needed for a present
or future task,
– is better entertained, more curious, more occupied with
interesting things,
– is happier, or enjoys in other ways a better quality of life.
130. Functions of an Information Retrieval System as
a Problem Solver
• determining information needs
• finding information
• applying information
131. Models for Online Retrieval Systems
• Traditional models
– A highly trained information specialist or librarian who
functions as a search intermediary, interpreting the
information requirements of the search requester and
translating these needs into a particular command language
and vocabulary of the search system being used.
• End-user systems
– This involves the development of user-friendly interfaces to
the conventional bibliographic retrieval system. Such
interfaces are designed for use by individuals who are not
computer specialists (or retrieval specialists in that case),
who have had only limited training, and who might reject a
system that is non-intuitive or complicated, or that uses
unfamiliar jargon to communicate.
132. Bibliographic Databases
• A bibliographic database consists of a set of records,
each of which uniquely characterizes a document in
terms of
– Author
– Publication source
– Subject matter
• The record serves as a representative of the actual
journal article or book
• It is intended to enable users to determine whether the
document is relevant to their information needs; and if
it is, where the document can be obtained.
133. Bibliographic Databases
• Online bibliographic-retrieval systems can rapidly
perform complex sorting and matching of hundreds of
thousands of file records using such techniques as
combinations of Boolean operators (“and”, “or”, and
“not”).
• Usually they are based on an electronic index to
journal or magazine articles, containing citations,
abstracts--and often either the full text of the articles
indexed, or links to the full text.
• Indexing plays an important role in this process.
Effective assignment of index terms
– Links documents that are related
– Discriminates finely among individual documents
134. Relational Databases
• A relation is defined as a set of tuples (records) that all have the
same attributes.
• This is usually represented by a table, which is data organized
in rows (i.e., the records) and columns (fields).
• In a relational database, all of the data stored in a column
should be in the same domain (i.e. data type).
• In the relational model, the tuples should not have any ordering.
• This means both that there should be no order to the tuples,
and that the tuples should not impose an order of the attributes.
• Put differently, neither the rows nor the columns should have an
order to them.
135. Indexed Databases
• Indexes can be created using one or more columns,
providing the basis for both rapid random lookups and
efficient ordering of access to records.
• The disk space required to store the index is typically
less than the storage of the table (since indexes
usually contain only the key-fields according to which
the table is to be arranged, and excludes all the other
details in the table).
• In a relational database an index is a copy of part of a
table.
136. Knowledge and skills should be acquired by
the researcher to achieve the following
objectives (aims):
• Most recent developments
• Most relevant hits
• Least amount of time consumed in the search
process
• Minimal cost
137. Retrieval Relevance
• It can be optimized by
–Minimizing the occurrence of false positives
(maximizing specificity)
–Minimizing the occurrence of false negatives
(maximizing sensitivity)
138. 05.05 World Wide Web
Because the WWW has proven to be very successful and because it is accessible to a
large public, many providers of on-line information are now making their information
available on a WWW service. An example is MEDLINE, a large database with
abstracts of all medical articles that appear in the international refereed medical
journals. It is also available as a WWW server called Internet Grateful Med. Via
ordinary WWW browsers, the MEDLINE database can be searched by entering
keywords. The retrieved abstracts are then formatted as WWW documents and sent to
the browser.
145. Endoscopic procedure. Physicians can select the images that they wish to store.
They can also add arrows and comments to the selected images for later
reference.
147. Dose distribution in tissue due to a single beam. The lines (isodose curves)
connect points with the same absorbed dose. (a) normal incidence (b) oblique
incidence
148. Reconstruction of a sagittal cross section and a few horizontal cross sections
149. Example of 3-D presentation of the chest after boundary detection, labeling,
shadowing and coloring of organs
151. Architecture for a clinical information system. Communication between the
modules is based on standardized protocols. This architecture lets applications
from multiple developers coexist. Integration takes place at the database level,
and some integration can also occur at the application level. A single results
review application can present data from disparate sources by accessing the
patient database.
152.
153. Network interconnecting systems in a department of cardiology, giving access to
a hospital information system (a), an ECG management system (b), a PACS for
coronary angiograms (c), a system for echocardiography (d), workstations (e),
and a CPR system (f).
161. A decision model may help in selecting the best features, and once the best
features are available, the decision model itself may be optimized. Observe the
similarity with detection: once we have a detector, a signal can be better
detected; when we know the signal, a better detector can be designed
162. Decision-support models in health care can be grouped into different categories.
The main categories are the quantitative (statistical) and the qualitative (heuristic)
decision-support models, which can also be further split into subcategories.
163. Example of a flowchart in the form of a computer program, based on similar
elementary decisions and logical expressions
164. Example of how clinicians may change their insight regarding a diagnosis. The
assignment of an expression of certainty should be expressed with each
diagnosis, together with the specification of the time period when the insight
applies.
166. Age distribution for boys and girls with childhood leukemia.
Generic Statistical Model
167. Hypothesis Testing
The number of outcomes (x patients with complications in a study of sample size
= n) can be modeled by a binomial distribution with parameters p and n, where p
denotes the known chance (probability) for untreated patients to experience the
complication.
The binomial distribution of x is denoted by the expression:
x ~ B(p, n).
In this context, x is also denoted as the test statistic. If p1 denotes the unknown
probability for complications in treated patients, one can define the null
hypothesis:
H0: p1 = p
and the alternative hypothesis:
H1: p1 # p
168. Hypothesis Testing
This means that under the null hypothesis one assumes that the therapy has no
effect (the probability of complications is equal to that known for untreated
patients), and under the alternative hypothesis one assumes that both
probabilities are different.
This type of alternative hypothesis is called two sided, because it makes no
assumption on a possible positive or negative effect of the treatment.
If one does not really have any definite reason to assume superiority of the
treatment under study (which will be true in most practical applications) one
should use a two-sided hypothesis.
If one has convincing reasons to assume superiority of the treatment under
study, one could also state a one-sided alternative hypothesis, for example,
H1: p1 < p
which is only looking for superiority of the treatment. In this case, the calculated
figures for the two-sided hypothesis presented above would have to be adjusted.
177. Distributions of systolic blood pressures of hypertensive and non-hypertensive
people for two hypothetical populations. Distributions for people in a population
survey (a), people in primary care (b), and people who visit a cardiac clinic (c)
are presented. For all three populations, the two distributions for hypertensive
and healthy people show considerable overlap. Decision thresholds are
indicated for all three populations. It can be seen that the choice of the decision
thresholds is different for each population (indicated by the shaded areas).
178. In (a) the two distributions of the primary population of Fig. 15.6b are presented,
but now with 10 different decision thresholds L1, L2, . . . L10. For each of these
decision thresholds the FP and FN percentages have been computed. These 10
combinations (FP, FN) have been plotted in the ROC curve (a) of Fig. 15.8. In (b)
the two distributions are closer together (i.e., more overlap between the
distributions) and in (c) the distributions are further apart. This is in Fig. 15.8
reflected in the ROC curves (b) and (c), respectively.
179. ROC curves of the populations of Fig. 15.7, with FP along one axis and FN (or 1-
CP) along the other axis. (See legends of Fig. 15.7.) The less the histograms
overlap, the better the ROC approaches the ideal point of (FP, FN) = (0, 0). The
more the histograms overlap, however, the more the ROC approaches the
diagonal line that runs from the point (FP, FN) = (100, 0) to the point (FP, FN) = (0,
100).
181. Illustration of Bayes' Rule. The distributions of the cell area for a collection of
lymphocytes (a) and for a collection of neutrophils (b) are shown. The prior
probability that an unclassified cell is a neutrophil (60%) is higher than its prior
probability of belonging to the class of lymphocytes (30%). When an area of 75
mm2 is measured, the posterior probability that the cell is a lymphocyte is higher.
182. Variability in symptoms between patients who possess the same disease. The
variability can be expressed in, for example, means and variances.
183. Symptoms or measurements, in short: features, determined for the same
disease category may be correlated and can also be expressed statistically. In
this diagram, observations for features f1 and f2 have been plotted.
184. Example of a training set that consists of only two classes, classes A and B, and
of which only two features, features f1 and f2, are known. (a) Clusters A and B are
fully separated, which is also indicated by the line between the clusters. (b) The
clusters are closer to each other and show some overlap. (c) A considerable
overlap is seen between clusters; panel d is identical to panel c, but the labeling
of the different elements of clusters A and B has been omitted. Two different
clusters can no longer be discerned, and it is clear that unsupervised clustering
techniques will also not be able to find a meaningful solution.
186. Table 18.5. Score Chart to Estimate the Probability of Spontaneous
Pregnancy in Subfertile Couples within one Year
Predictor Infertility score
Age of female (years)
21 to 25
26 to 31
32 to 34
35 to 37
38 to 40
0
2
4
10
16
Duration of infertility (years)
1
2
3 to 4
5 to 6
> 7
0
2
4
8
12
187. Table 18.5. Score Chart to Estimate the Probability of Spontaneous
Pregnancy in Subfertile Couples within one Year
Predictor Infertility score
Female infertility
Secondary
Primary
0
7
Fertility problems in male's family
No
Yes
0
5
Post-coitum test
Progressive
Nonprogressive
Negative
0
10
20
Motility (%)
> 60
40 - 59
20 - 39
0 - 19
0
3
7
10
Prognostic index (encircle relevant scores and add)
188. Relation between the prognostic index in Table 18.5 and the predicted pregnancy
rate at 1 year.