COMP3330: History Assignment

Artificial Intelligence Applications In Medicine

Conclusion & Timeline

This page contains:

the conclusion of the report

a timeline to illustrate the significant developments in the history of AI in medicine

Conclusion

In the past few years, AI in medicine has slowly regained its popularity. Medical AI systems, and the technologies for developing these systems, have been proven in a vast variety of mainstream medical applications.

Furthermore, as massive encrypted health information becomes available online for medical researchers, sophisticated computer techniques will be needed to make use of the data in a meaningful way. Indeed in the near future, many medical applications will need to include intelligent software components simply to remain competitive.

Many areas of current AI research are dedicated to developing systems that utilise a combination of the types of systems presented above (e.g. neural networks and fuzzy logic, expert systems with integrated intelligent DSS, etc). In the future, these combinations of the various AI technologies will continue to be explored, and may be combined to form an integrated system.

An example of an integrated medical AI system is given below.

Figure 3: A fully integrated medical AI application

As the feasibility of such systems is increasingly demonstrated, there will be increased demand for these integrated intelligent systems, which will enhance the performance of medical devices and help users to interpret complex outputs. Overall, the use of artificial intelligence within medical applications will continue to improve the standard of medical care, while at the same time helping to contain the expanding costs of healthcare providers.

Timeline

This timeline is provided to illustrate the significant developments in the history of AI in medicine.

Table 1: Applications of AI in medicine.

Time Period	Event
Mid 1950s	A frequent goal of early attempts at clinical expert systems is to virtually replace the physician with a Greek oracle model of clinical decision making. The object of this line of thinking is to create a "doctor in a box", capable of querying the physician or medical technician regarding a patient's symptoms and generating a diagnosis.
Late 1960s	Early expert systems sparked considerable excitement in the filed of medicine, and resulted in a high level of expectation in the late 1960s.
1970	Dr William B. Schwartz publishes an influential paper about AI in medicine in the New England Journal of Medicine. As a result, many scientists are attracted to study the applications of computer science in medicine.
1970	Harvard Medical School and MIT establish a joint institution to foster the development of health related education, research and service. The new division was called Health, Sciences, and Technology (HST) Among many programs, HST offered training in medical informatics, a field closely related to AI in medicine.
1975	Development of MYCIN expert system, a rule-based program for diagnosis of bacterial infections in the blood. (Stanford)
1975	Development of INTERNIST expert system, a diagnostic aid that combines a large database of disease/manifestation associations with techniques for problem formulation. (Pittsburgh)
1976	Development of CASNET (Causal-Associated Network) expert system, which uses physiological models for the diagnosis and treatment of eye disease. (Rutgers Research Resource)
1977	Development of PUFF expert system, for automatic interpretation of pulmonary function tests. (Pacific Presbyterian Medical Centre, San Francisco)   The PUFF system has been sold in its commercial form to hundreds of sites worldwide and is still in use today.
Late 1970s	Development of PRESENT-ILLNESS expert system, to diagnose kidney diseases. (MIT)   It employs a computer model that is more complicated than the rule-based systems being used in other projects, a "frame-based" model in which living things are represented by a frame-like structure within the program.
1979	Establishment of the American Association for Artificial Intelligence (AAAI).
1981	Development of ABEL expert system, a program that uses multi-level pathophysiologic models for diagnosis of acid-base and electrolyte disorders. (MIT)
Early 1980s	AI in medicine is a largely US-based research community. Work originates out of a number of campuses, including MIT, Pittsburgh, Stanford and Rutgers.
1983	Development of PDS expert system. (CMU)
1983	Development of MED1 expert system. (Kaiserslautern)
1984	Clancey and Shortliffe provide the following definition of medical AI: "Medical artificial intelligence is primarily concerned with the construction of AI programs that perform diagnosis and make therapy recommendations. Unlike medical applications based on other programming methods, such as purely statistical and probabilistic methods, medical AI programs are based on symbolic models of disease entities and their relationship to patient factors and clinical manifestations." [Clancey, 1984] Much has changed since then, and today this definition would be considered narrow in scope and vision. Today, the importance of diagnosis as a task requiring computer support in routine clinical situations receives much less emphasis.
1985	Development of MED2 expert system. (Kaiserslautern)
Mid 1980s	Establishment of an organization called the 5th Generation Project in Japan. One of its goals was to facilitate communication among medical AI scientists around the world.
Mid 1980s	The development of expert systems suffered from a few drawbacks at this time: some expert systems could not operate as well as the experts who supplied them with knowledge most of the expert systems had to be run on (costly) LISP machines the LISP machines could not be connected to a network These limitations hindered the development of expert systems as commercial applications at the time.
1986	Establishment of European Society for Artificial Intelligence in Medicine (AIME) in Europe.
1989	Development of MUNIN expert system for diagnosing neuromuscular disorders.
Early 1990s	Failure of researchers to present satisfactory systems resulted in reduced funding from the government and venture capitalists in the early 90s. AI scientists often refer to the period of 1987-1994 as the "AI winter."
1991	Development of PEIRS (Pathology Expert Interpretative Reporting System) expert system, to generate pathology reports. PEIRS reported on a variety of pathology measures, with an overall diagnostic accuracy was about 95%.
1993	Development of GermWatcher AI laboratory system. This system checks for hospital-acquired (nosocomial) infections, which represent a significant cause of prolonged inpatient days and additional hospital charges.   Microbiology culture data from the hospital's laboratory system are monitored by GermWatcher, using a rule base containing a combination of national criteria and local hospital infection control policy.
Mid 1990s	Research in medical AI changes its focus to new areas including: getting more data online building better information infrastructure using machine learning techniques to make decisions learning and developing better easy-to-use programs suitable for health professionals A central part of all these initiatives was the creation of Electronic Medical Records (EMR), which serve as the central clinical repository of information pertaining to patient care.
1997	The (American) National Library of Medicine awards contracts to a variety of health care organisations across the country to investigate innovative uses of the national information infrastructure for health care, including tele-medicine and information sharing.
Near Future	Completion of the human genome project will be likely to lead to the following applications of AI in medicine: repositories of knowledge created by AI systems in hospitals across the world will be connected, allowing them to share all the information collected. This information will be aggregated using distributed computing and subsequent analysis by the AI system; it will draw inferences from data mining applications that will develop patterns based on this aggregated data. expert systems and neural networks will be used to determine medically valuable patterns genetic algorithms will be used to ensure that the systems continue to learn The results of this analysis will be fed back to individual hospitals' AI inference engines to allow their AI software to analyse each patients data in reference to current patterns of diseases, surgery complication, medicine complications with certain genome types, etc.