Artificial Intelligence in the Clinical Laboratory
How does it work?
By: Ralph Anderson
Introduction
I intend to define Artificial Intelligence, and how, when
used in the clinical laboratory, AI can be applied to the field of Psychology.
I will also draw a conclusion from this research, including why I choose
AI.
Summery
Sophisticated software that can monitor and control processes and interpret
information, now readily available, can be used "...to incorporate intelligence
into advanced analytical instrumentation. This form of machine intelligence
is called Artificial Intelligence." (Place, Truchaud, Ozawa, Pardue, and
Schnipelsky, 1994) The following are points I feel necessary to illustrate
what AI is and how it works:
Automaton: an automaton is a machine that is capable of repetitive
tasks and uses mechanical forms of feedback. This is one of the earliest
forms of AI, made popular during the Industrial Revolution. With
the introduction of Electronics and Digital Computing, increased memory
capacity, sensory technology, and the modern theories of control systems
(like AI), it is possible to have robotic arms with full function, space
rovers that determine their terrain and adjust accordingly, and anything
else you can dream up. This includes common forms of AI, such as
computer Chess.
"Any form of automation or decision making system can be regarded very
simply as an input giving rise directly or indirectly via a process system
to an output, which then gives rise to some feedback or control." (Place
et
al., 1994)
Intelligence: Through the birth of programmable software,
the introduction of flexible operational processes and feedback loops to
the AI system allowed it the ability to become "Intelligent". The
definition of intelligence states "An individuals overall capacity to think
rationally, act purposefully, and deal effectively with the environment."
(Coon, 1997), and AI is reliant on the mechanisms of learning. AI
is able to 'learn' by the accumulation of knowledge.
Knowledge: Today, AI and its technologies are generally referred to as "Knowledge Engineering". The computer software that encompasses this knowledge is called an "Expert System". Expert systems, through specially programmed, accumulated knowledge bases, are able to make decisions. Drawing on the human expertise that has been meticulously gathered, interpreted, and entered into the system makes these decisions. Some advantages to this are:
a) The ability to draw on the expertise of many professionals, simultaneously, to make clear, knowledgeable, logical conclusions.
b) The cost of the information is considerably lower; the information is
already
in the computer, so there is no need to tie up dozens of specialized individuals
for their opinion.
Inference Engine: "The inference engine implements the problem
solving strategies built up by the knowledge engineer...and is independent
of the contents of the knowledge base." (Place et al., 1994)
By using an IF-THEN statement, it can determine if a situation is true.
IF-THEN lines state that IF a condition is True\False, THEN perform some
action. Sometimes, however, the information is not completely accurate.
This is most likely to occur in a situation where the amount of knowledge
present is insufficient to draw a steadfast conclusion.
Neural Networks: Since decisions are based almost entirely on what
is learned by experience, the AI gurus had to devise a different approach,
one that sorts out pattern classification and functional synthesis.
Like the human brain, the "...neural network acts as an associative memory,
...good at generalizing from specific examples..."(Place et al., 1994)
Neural Networks have to be trained, rather than programmed, by processing
time-dependent spatial patterns. Future AI systems may use a combination
of the refined neural network to solve a problem, and an expert system
to explain the logical solution provided.
Applications
Medicinal Purposes: With the explosion of knowledge and data in
the medical field, the use of a computer-supported medical decision-maker
may be the solution. "Many medical problems, especially those related
only to laboratory data, can be solved using conventional algorithms as
they contain data of known imprecision. It has been suggested that
only solutions requiring between10 min and 3 hrs of clinician time are
suitable for expert systems (Frenzel, LE, 1987)"
Embedded Systems: At many levels, embedded knowledge-based systems
are used in laboratory instrumentation. There are applications for
failure, detection, testing and maintenance. According to Groth and
Moden, 1991, they describe a system providing real time quality controller
multi-test analyzers using a relational database management system.
Vision: A machines vision is able to detect minute differences
that a human cannot. This is useful in morphologic architecture.
User Interfaces
The success of an AI system relies on the user interface, such as:
Keyboard, speakers, microphones, touch pads, the weather and barometric
pressure.
Conclusion
We have looked at ways in which the development of programmed software
and its flexibility has allowed clinical laboratory systems to become intelligent.
We have also looked at how the neural network learns. This is based
on the methods of human learning. I wonder haw soon it will be before
a machine will act, look and think like a human. What then?
If I was to do an experiment of AI, I would be most interested in the neural
networks, and how to improve their learning capability. Someday I hope
to be working with cybernetic organisms, and possibly teaching them, and
helping them move beyond the "Dr. Spock" state of consciousness.
Milestones
in AI: To see the infancy of AI though
it’s adolescence to present day, visit " Milestones in the
Development of Artificial Intelligence" ftp.cs.cmu.edu:/user/ai/pubs/faqs/ai/timeline.txt
Glossary
of Terms (Parent document is top of "FAQ: Artificial Intelligence
Questions & Answers 1/6 [Monthly posting]" http://www.cis.ohio-state.edu/hypertext/faq/usenet/ai-faq/general/part1/faq-doc-3.html)
AI - Artificial
Intelligence: "Any
artificial system (often a computer program) that is capable of human-like
problem solving or skilled responding." (Coon, 1997)
Admissibility: An admissible search algorithm is one that is guaranteed
to find an optimal path from the start node to a goal node, if one exists.
In A* search, an admissible heuristic is one that never overestimates the
distance remaining from the current node to the goal.
Acquired Strategies: "Learned tactics for swiftly solving the problems
encountered in one's area of expertise." (Coon, 1997)
Autonomous Vehicles: Unifies Robotics, Machine Vision, Machine
Learning, Intelligent Control, Planning. (Taken from Parent document, top
of "FAQ: Artificial Intelligence Questions & Answers 1/6 [Monthly posting]"
http://www.cis.ohio-state.edu/hypertext/faq/usenet/ai-faq/general/part1/faq-doc-10.html)
Case-based Reasoning: Technique whereby "cases" similar to the
current problem are retrieved and their "solutions" modified to work on
the current problem.
Data Mining: Also known as Knowledge Discovery in Databases (KDD)
was been defined as "The nontrivial extraction of implicit, previously
unknown, and potentially useful information from data" in Frawley and Piatetsky-Shapiro's
overview. It uses machine learning, statistical and visualization
techniques to discover and present knowledge in a form, which is easily
comprehensible to humans.
Expert Systems: "Computer programs designed to respond as a human
expert would; programs based on the knowledge and rules that underlie human
expertise in specific topics." (Coon, 1997)
Fuzzy Logic: In Fuzzy Logic, truth-values are real values in the
closed interval [0.1]. The definitions of the Boolean operators are extended
to fit this continuous domain. By avoiding discrete truth-values, Fuzzy
Logic avoids some of the problems inherent in either-or judgments and yields
natural interpretations of utterances like "very hot". Fuzzy Logic has
applications in control theory.
Intelligence: "An individuals overall capacity to think rationally,
act purposefully, and deal effectively with the environment." (Coon, 1997)
Knowbots/Infobots and Intelligent Help Desks: Unifies Information
Retrieval, Reasoning, Intelligent User Interfaces, Qualitative Reasoning.
(Taken from Parent document, top of "FAQ: Artificial Intelligence Questions
& Answers 1/6 [Monthly posting]" http://www.cis.ohio-state.edu/hypertext/faq/usenet/ai-faq/general/part1/faq-doc-10.html)
Nonlinear Planning: A planning paradigm, which does not enforce
a total (linear) ordering on the components of a plan.
Organized Knowledge: "Orderly and highly refined information about
a particular topic or skill." (Coon, 1997)
Strong AI: Claim that computers can be made to actually think,
just like human beings do. More precisely, the claim that there exists
a class of computer programs, such that any implementation of such a program
is really thinking.
Validation: The process of confirming that one's model uses measurable
inputs and produces output that can be used to make decisions about the
real world.
Verification: The process of confirming that an implemented model
works as intended.
Weak AI: Claim that computers are important tools in the modeling
and simulation of human activity.
References