International Dental Journal (2010) 60. 269-272

Artificial intelligence: contemporary applications and future compass Sunali Khanna

Department of Oral Medicine and Radiology, Nair Hospital Dental College, Mumbai, India

The clinical use of information technology in the dental profession has increased substan- tially in the past 10 to 20 years. In most developing countries an insufficiency of medical and dental specialists has increased the mortality of patients suffering from various diseases. Employing technology, especially artificial intelligence technology, in medical and dental application could reduce cost, time, human expertise and medical error. This approach has the potential to revolutionise the dental public health scenario in developing countries. Clinical decision support systems (CDSS) are computer programs that are designed to provide expert support for health professionals. The applications in dental sciences vary from dental emergencies to differential diagnosis of orofacial pain, radiographic interpre- tations, analysis of facial growth in orthodontia to prosthetic dentistry. However, despite the recognised need for CDSS, the implementation of these systems has been limited and slow. This can be attributed to lack of formal evaluation of the systems, challenges in developing standard representations, cost and practitioner scepticism about the value and feasibility of CDSS. Increasing public awareness of safety and quality has acceler- ated the adoption of generic knowledge based CDSS. Information technology applications for dental practice continue to develop rapidly and will hopefully contribute to reduce the morbidity and mortality of oral and maxillofacial diseases and in turn impact patient care.

Keywords: Artificial intelligence, clinical decision support system, diagnosis, dental - applications

Clinical Decision Support Systems (CDSS)

CDSS are interactive computer programs, which are designed to assist physicians and other health profes- sionals with decision making tasks. A worhng defini- tion has been proposed by Dr. Robert Hayward of the Centre for Health Evidence: “Clinical Decision Support systems h k health observations with health knowledge to influence health choices by clinicians for improved health care”. This definition has the advantage of sim- plifylng chical decision support to a functional concept.

The basic components of a CDSS include a 4namic ( m e d d ) knowledge base and an inferencing mechanism (usually a set of rules derived from the experts and evidence-based medicine) and implemented through medcal logic modules based on a language such as Arden syntax. It could be based on expert systems or artificial neural networks or both (connectionist expert systems).

CDSS are computer programs that are designed to provide expert support for health professionals making

clinical decisions. These systems use embedded clinical knowledge to help health professionals analyse patient data and make decisions regarding diagnosis, prevention and treatment of health problems. Examples of such systems can be found in several disciplines in health care: dentistry, medicine and pharmacy, among others’.

CDSS applications may be stand alone systems, or they may interact with other tools such as an electronic dental record, an order entry system, or a radiology system. They may decline a recommendation for the patients’ treatment and offer options on future evalua- tion. CDSS may also generate alerts regarding potentially dangerous conditions for a patient (drug allergies), or they may remind clinicians of routine tasks such as more frequent screening for oral cancer in a smoker, for periodontal disease in a patient with diabetes or even to perform tasks such as the use of prophylactic antibiotics when appropriate. The varied applications include radiology systems and patient education tools which may provide dentists with additional support2.

0 2010 FDlNVorld Dental Press doi:l0.1922/IDJ-2422Khanna04 0020-653911 0104269-04

270

Historical perspective

The use of computers to assist health professionals in their activities has been studed since the 1950s. Initial work was focused on the development of diagnostic systems3. Since then, researchers have applied dif- ferent methods to provide clinical applications with knowledge. More recent work on CDSS has focused on integration of these applications with clinical da- tabases. These integrated systems take advantage of data already recorded for other purposes in order to avoid redundant data entry in the provision of alerts and reminders. These CDSS may monitor data in large health care organisations or may be part of an electronic patient record installed in a single c h c a l office or c h c . Research has also been done on eliciting patients’ pref- erences for therapeutic options, whch can help health care professionals to gain a better understandmg of their patients’ perspectives4.

Methodology of CDSS

Most CDSS have four basic components: Inference En- gine (IE), Knowledge Base (I(B), Explanation Module and Worhng Memory (Figzrre I).

The Inference Engine (IE) is the main part of any such system, containing the knowledge about the pa- tient from which to draw conclusions regardlng certain conditions. The knowledge used by IE is represented in the knowledge base and tools that have been cre- ated to facditate the acquisition and elicitation of t h s knowledge. The collected patient data may be stored in a database or may exist in the form of a message and is known as the worhng memory. Patient data may include demographics (i.e. date of birth, gender), al- lergies, medications in use, previous dental or medical problems and other information. The last component, the Explanation Module is not present in all CDSS. This module is responsible for composing justifications for the conclusions drawn by the IE in applying the knowl- edge in the KB against patient data in working memory.

a- n

Figure 1. The basic components of a CDSS

An insight into dental applications

This can be grouped in seven subareas of dentistry:

Dental emergencies and trauma Orofacial pain (differential diagnosis) Oral medicine (management of oral dseases in neck and head) Oral radiology (interpretation of radiographic lesions and automated interpretation of dental radlographs) Orthodontics (analysis of facial growth, landmark identification on cephalometric radlographs, and treatment planning) Pulpal diagnosis and restorative dentistry (remov- able partial denture design).

More than 85% of dentists use computers in their offices. Although most of this growth ishue to the use of patient accounting, bdling and scheduling systems, the clinical use of information technology in the dental profession has increased substantially in the past 10 to 20 years. Despite the recognised need for CDSS, the implementation of these systems has been limited, for which there are several reasons. Lack of formal evalu- ation of these systems, challenges in developing stand- ard representations, lack of studes about the decision making process, the cost and difficulties involving the generation of knowledge base, and practitioner scepti- cism about the value and feasibility of decision support systems are among others.

Generally CDSS are proliferating as fragmented and isolated systems with a few clinic or hospital-wide exceptions in academic centres. In parallel, the public awareness of safety and qualtty has accelerated the adop- tion of generic knowledge based CDSS. Information technology application for dental practice continues to develop rapidly and wdl hopefully contribute to the re- moval of part of the data entry barrier. Hybrid systems attempt to overcome these drawbacks by combining both deducting rules and probabilistic reasoning in the same CDSS. They use features of several or all the pre-

Explanation 1 Mod; 1 ym Memory

International Dental Journal (2010) Vol. 60/No.4

271

viously described systems along with heuristics to assist clinicians in mahng decisions.

Lessons learned from these systems revealed the feasibhty of encodng clinical knowledge and helped researchers to clarify both the strengths and limitations of knowledge representation approached. There is a gradual change in attitudes and increasing acceptance of computer decision tools by healthcare professionals.

However, this enthusiasm can &minish if research- ers do not ensure that the products of their research respond to real world needs and are sensitive to the logistical requirements of the practice settings in which clinicians work’.

However, there are sull more challenges to be over- come. The future of CDSS depends on the adoption of evidence-based practice, progress in developing use- ful programs, the adoption of standards to allow inter operabhty, the reduction of logistical barriers to imple- mentation, understanlng of the complex and changmg nature of clinical knowledge and proper validation of the programs. Additional challenges are related to the legal implications inherent to the development and use of such innovations.

Technical challenges and future prospects

Most approaches to computer assisted &agnosis have, unul the past few years, been based on one of three strategies, statistical pattern matching or probabhty theory. All three techniques have been successfully ap- plied to narrow medical domains, but each has serious drawbacks when applied to broad areas of clinical medicine.

Flowcharts quickly become unmanageably large. Further, they are unable to deal with uncertainty, a key element in most serious &agnostic problems. Probabi- listic methods and statistical pattern matching typically incorporate unwarranted assumptions, such as that the set of &seases under consideration is exhaustive, that the diseases under suspicion are mutually exclusive or that each clinical findmg occurs independently of all others.

Artificial intelligence is a study to emulate human intelhgence into computer technology. It provides for examination, organisation, representation and catalogu- ing of me&cal knowledge. It also offers a content rich discipline for future scientific mehcal specialty. Even though the system is equipped with human knowledge, the system wdl never replace human expertise as hu- mans are required to frequently monitor and update systems knowledge. Examples of some systems are MYCIN, CASMET and Internist - I etc. MYCIN is applied to diagnose certain antimicrobial infection and treatment recommendation. CASMET (casual asso- ciational networks) is for building expert systems for diagnosis and treatment of disease^^,^.

Fuzzy loglc is another branch of artificial intelhgence technique. It deals with uncertainty in knowledge that simulates human reasoning in incomplete or fuzzy data. Similarly, neural network is a powerful artificial intell- gence technique that has the capabhty to learn a set of data and constructs matrixes to represent the learning patterns. It is a network of many simple processors or units that simulate the function of the human brain to perform tasks as a human does. Central databases over the worldwide web enable information sharing, collabo- ration between me&cal practitioners, online discussion and online treatment and diagnosis’. Much effort has been put forth by medical institutions and software companies to produce viable CDSS to cover all aspects of clinical tasks. However, the complexity of clinical workflows and the hgh demands on staff care must be taken by the institution deploying the support system to ensure that it becomes a fluid and integral part of the workflow. Two sectors of the health care domain in which CDSS have had a large impact are pharmacy and accountancy.

Clinical decision support systems have been termed active knowledge systems, which use two or more items of patient data to generate case specific advice. This implies that CDSS is simply a lagnostic decision system that is focused on using knowledge management in such a way as to achieve clinical advice for patient care based on some number of items of patient data’.

Clinical decision support systems face steep technical challenges in a number of areas. Biological systems are profoundly complicated and a clinical decision may uti- lise an enormous range of potentially relevant data, for example an electronic evidence- based medicine system may potentially consider a patient’s symptoms, medical history, family hstory and genetics as well as hstorical and geographical trends of &sease occurrence, and published clinical data on medical effectiveness when recommending a patient’s course of treatment.

Programs using artificial intelhgence techniques have several major advantages over those using more trah- tional methods. These programs have a greater capacity to quickly narrow the number of &agnostic possibilities, they can effectively use pathophysiologic reasoning and they can create models of specific patient illness. Such systems can even capture the complexities created by several diseases states that interact and overlap.

In most developing countries insufficiency of me&- cal specialists has increased the mortality of patients suf- fering from various diseases. General medical practition- ers may not have enough expertise to deal with certain high risk lseases which could be cured at an early age; employing technology especially artificial intelligence technology in medical application could reduce cost, time, human expertise and me&cal error. Advances in h s field have the potential to impact medical and dental public health scenarios in developing countries and in turn patient care.

Khanna: Artificial intelligence

272

1.

2.

3.

4.

5.

6.

7. 8.

References

Musen MA, Shahar Y, Shortliffe EH. Clinical decision - support systems. In: Shortliffe EH, Pereault LE, Widerhold G, Fagan LM, eds. Medical informatics: Computer applications in health care and biomedicine. 2"d Ed. Pp 573-609. New York: Springer- Verlag, 2001. EA. Mendonca. Clinical Decision Support *stems: Perqectives in Dentistry. pp 589- 596. Miller RA. Medical diagnostic decision support systems - past, present and future: a threaded bibliography and brief commentary. J and M Mid I n j m Assoc 1994 1: 8-27 . Ruland CM, Bakken S. Developing, implementing and evaluating decisions support systems for shared decision making in patient care: a conceptual model and care illustration. JB Combed Inform

Proceedings of National Conference on Research and develop- ment in computer science in its applications University Pertanian Malaysia: Kuala Lumpur. pp 220-224. Hoong NK. Medical information Science Framework in Potential, International Seminar in exhibition, computerization for develop- ment - the research challenge. pp 191-198. University Pertanian Malaysia: Kuala Lumpur, 1988. Shorthffe EH. Standford edu/pubs/abstracts by author Berner, Eta S, ed. Clinical Decision Support Systems New York,

NY: Springer 2007

2002 35: 313-312.

Correspondence to: Dr Sunali Khanna, Department of Oral Medi- cine and Radiology, Nair Hospital Dental College, Mumbai 400 008, India. Email: sunali3011@yahoo.com

International Dental Journal (2010) Vol. 601No.4