Monthly Archives: November 2013

CPU platforms

Embedded processors can be broken into two broad categories: ordinary microprocessors (^P) and microcontrollers (p, C), which have many more peripherals on chip, reducing cost and size. Contrasting to the personal computer and server markets, a fairly large number of basic CPU architectures are used; there are Von Neumann as well as various degrees of Harvard architectures, RISC as well as non-RISC and VLIW; word lengths vary from 4-bit to 64-bits and beyond (mainly in DSP processors) although the most typical remain 8/16-bit. Most architectures come in a large number of different variants and shapes, many of which are also manufactured by several different companies.

Ready made computer boards

PC/104 and PC/104+ are examples of available ready made computer boards intended for small, low-volume embedded and ruggedized systems. These often use DOS, Linux, NetBSD, or an embedded real-time operating system such as MicroC/OS-II, QNX or VxWorks.

In certain applications, where small size is not a primary concern, the components used may be compatible with those used in general purpose computers. Boards such as the VIA EPIA range help to bridge the gap by being PC-compatible but highly integrated, physically smaller or have other attributes making them attractive to embedded engineers. The advantage of this approach is that low-cost commodity components may be used along with the same software development tools used for general software development. Systems built in this way are still regarded as embedded since they are integrated into larger devices and fulfill a single role. Examples of devices that may adopt this approach are ATMs and arcade machines. and which contain code specific to the application.

ASIC and FPGA solutions

A common configuration for very-high-volume embedded systems is the system on a chip (SoC) which contains a complete system consisting of multiple processors, multipliers, caches and interfaces on a single chip. SoCs can be implemented as an application-specific integrated circuit (ASIC) or using a field-programmable gate array (FPGA).


Si ——————————— GaAs


Note Examples are for N ~ 1*1017 cm-3 and y = 4 V There is also a quick way to estimate the depletion width For most semiconductors, for tj/ = 1 VandA^ lxlO15 cm-3, ^happens to be around 1 fim Since Wd<x {tf//N)^{Eqs (El 19)—(El 20)), we can scale from this set of numbers For example, with N= MO17 cm-3, Wd becomes 0 1 jim Changing y = 4 V would then increase the fVdto0 2 jini



Soekris net4801, an embedded system targeted at network applications.

1. Embedded systems are designed to do some specific task, rather than be a general-purpose computer for multiple tasks. Some also have real-time performance constraints that must be met, for reasons such as safety and usability; others may have low or no performance requirements, allowing the system hardware to be simplified to reduce costs.

2. Embedded systems are not always standalone devices. Many embedded systems consist of small, computerized parts within a larger device that serves a more general purpose. For example, the Gibson Robot Guitar features an embedded system for tuning the strings, but the overall purpose of the Robot Guitar is, of course, to play music. Similarly, an embedded system in an automobile provides a specific function as a subsystem of the car itself.

3. The program instructions written for embedded systems are referred to as

Firmware, and are stored in read-only memory or Flash memory chips. They run with limited computer hardware resources: little memory, small or non-existent keyboard and/or screen.

User interfaces


Embedded system text user interface using MicroVGA

Embedded systems range from no user interface at all — dedicated only to one task — to complex graphical user interfaces that resemble modern computer desktop operating systems.

Simple systems

Simple embedded devices use buttons, LEDs, and small character or digit-only displays, often with a simple menu system.

In more complex systems

A full graphical screen, with touch sensing or screen-edge buttons provides flexibility while minimising space used: the meaning of the buttons can change with the screen, and selection involves the natural behavior of pointing at what’s desired.

Handheld systems often have a screen with a "joystick button" for a pointing device.

Many systems have "maintenance" or "test" interfaces that provide a menu or command system via an RS-232 interface. This avoids the cost of a display, but gives a lot of control. Most consumers cannot assemble the required cables, however.

The rise of the World Wide Web has given embedded designers another quite different option: providing a web page interface over a network connection. This avoids the cost of a sophisticated display, yet provides complex input and display capabilities when needed, on another computer. This is successful for remote, permanently installed equipment such as Pan-Tilt-Zoom cameras and network routers.

Medical Terminology and Diagnosis Using Knowledge Bases




Classification and Coding Systems


An Electronic Medical Encyclopedia at Your Fingertips: Knowledge Bases and Diagnosis

Peter L. M. Kerkhof


Problems Related to Medical Terminology

Medwise Working Group


Solution for Discrepancies

This man with two blue hands and abnormal blood gases and a negative chest X-ray. And high blood pressure. ..(one physician) said he didn’t know what the hell was going on. Would I take a look at him? Just a curbside consultation—that’s all. Just an opinion. I said O. K.—be glad to. It was entirely appropriate.

I was a physician of record.”

From Two Blue Hands, by Berton Roueche


Computers in medicine assist the process of communication and support the integration of information. In the above example given by Roueche, the consulting physician could be replaced by a computer system that stores and handles information that was previously obtained from a number of experts in various medical specialties. Ideally, the victim described here carries a patient data card which, in an emergency case, presents valuable information to the physician. The Medical Records Institute [1981] is actively engaged in the development of an electronic patient record.

Computer-mediated reasoning systems have been described by Williams [1982]: Knowledge Base (KB) systems are used for interpretation of data about a specific problem, in the light of knowledge represented in the KB, to develop a problem specific model and then to construct plans for problem solution. The KB contains the descriptive or factual knowledge pertaining to the domain of interest. Candidate hypoth­eses are then derived through some pattern matching system. Next, a reasoning “engine” (also termed inference machine) carries out the manipulation specified to reach a decision. Knowledge obviously can be represented in several ways: it can be characterized by symbols, plain words, and definitions along with their interrelations. Knowledge may be expressed by spoken or written words, flow charts, (math­ematical) equations, tables, figures and so forth, depending on the level of abstraction.

Aspects of language and text interpretation are central issues in artificial intelligence. Language is regarded as a metaphor of thought. When it provides the primary abstraction of knowledge about reality, it seems reasonable to expect that a powerful abstraction of language should also provide a powerful representation of knowledge. Various strategies have been explored: semantic networks offer a versatile tool for representing knowledge of virtually any type that can be captured in words by employing nodes (representing things) and links (referring to meaningful relationships between things), thus expressing causal, temporal, taxonomic and associational connections. Other approaches (such as frame-systems and production systems) have also been investigated. Conceptual Graphs (ANSI X3H4) [Sowa, 1984] are an emerging standard for knowledge representation, and the method is particularly suited to the repre­sentation of natural language semantics. Weed [1991] has advanced the idea of “knowledge coupling”

E., a combination of a computerized medical record and decision-support software.

An ideal medical KB must be comprehensive (while integrating text, graphics, video, and sound), accurate and verifiable, easily accessible where doctors see patients, and the system should be adaptable to their own preferred terms or abbreviations [Wyatt, 1991]. Future developments will certainly include the use of artificial neural networks [Baxt, 1991], and examples realized thus far include myocardial infarction, diabetes mellitus, epilepsy, bone fracture healing, appendicitis, dermatology diagnosis, and EEG topography recognition.

Currently, most internationally available textbooks of (internal) medicine and medical dictionaries are retrievable on CD-ROM. Lexi-Comp’s Clinical Reference Library is an example, featuring user-defined hyperlinks. A variety of titles are available as digital books on data cards suitable for palm computers, such as the Pocket PDR. In addition, the Internet offers numerous sites with relevant information [Smith and Edwards, 1997]. The September 1997 issue of the journal, MD Computing, focuses on medical resources on the Web.

This chapter addresses several approaches employed to represent medical definitions and knowledge. First, a survey will be presented on available coding systems, and next an overview is given of current KB systems which primarily refer to the process of establishing a diagnosis. Finally, an anthology con­cerning problems related to medical terminology is presented, along with directions for obtaining a solution.

Classification and Coding Systems

With the exception of one British system, all classification or coding systems have been developed in the USA. This survey lists all projects along with some of their characteristics.

The International Classification of Diseases (ICD) [International Classification of Diseases, 1994] system entered its tenth version, succeeding the ninth edition released in 1978. With more than 8000 codes, it is applied worldwide for classifying diagnoses and also permits diagnosis related group (DRG) assignment employed for billing and reimbursement purposes.

Systematized Nomenclature of Medicine (SNOMED) [Systematized Nomenclature of Medicine, 1993] offers a structured nomenclature and classification for use in human as well as in veterinary medicine. It consists of eleven modules with multiple hierarchies covering about 132,600 records, with a printed as well as a CD-ROM version available.

Physicians’ Current Procedural Terminology (CPT) [Physicians’ Current Procedural Terminology, 1992] provides uniform language for diagnostic as well as surgical and other interventional services. This system is currently in its fourth edition which is distributed by the American Medical Association (AMA), and has been incorporated in the Medicare program.

Medical Subject Headings (MeSH) [1993] is a systematic terminology hierarchy which is used to index the MEDLINE medical publications system. It forms the standard for coding keywords related to the contents of articles in the field of medicine. The system is updated yearly.

The National Library of Medicine (NLM) in 1986 started a project called Unified Medical Language System (UMLS) [Lindberg et al., 1993]. The project aims to address the fundamental information access Problem caused by the variety of independently constructed vocabularies and classifications used in different sources of machine-readable biomedical information. The UMLS approach will be to compen­sate for differences in the terminologies or coding schemes used in different systems, as well as for differences in the language employed by system users, rather than to impose a single standard vocabulary on the biomedical community. The Metathesaurus contains 311,000 terms and maintains cross reference to CPT [Physicians’ CPT, 1992], ICD, SNOMED [SNOMED, 1993] and other indexes. NLM encourages broad experimentation with the fourth edition, and the tools are (with certain provisions) available on the basis of a one-year agreement. The complete files are distributed on four CD-ROMs, while the application programs are limited to browsing tools for the metathesaurus.

Gabrieli [1992] constructed a computer-oriented medical nomenclature based on taxonomic princi­ples. His system covers 150,000 preferred terms and a similar number of synonyms. The partitioning method employed for medical classification readily permits replacement of English names with terms of any other language, thus creating the perspective of a world-wide standard.

Read from the U. K. designed a classification for various computer applications [Chisholm, 1990]. It is designed in accordance with six key criteria: to be comprehensive, hierarchical, coded, computerized, cross-referenced, and dynamic. The system is closely connected with the British National Health Service (NHS), and includes 100,000 preferred terms, 250,000 codes, and 150,000 synonyms.

An Electronic Medical Encyclopedia at Your Fingertips: Knowledge Bases and Diagnosis

In this era, most information (as stored in written documents or computer records) consists of natural language text. Traditional retrieval systems use sets of keywords (index terms). It is called a full text retrieval system, when all (nontrivial) words of that text are indexed for future identification during search procedures. Information requests can then be formulated with the aid of Boolean operators (namely: and, or, not). Synonymous terms can be specified by employing the or-operator. The fuzzy-set retrieval model still uses Boolean logic but refines the query process by assigning different weight factors to the individual index terms. As an attractive alternative, the vector space processing system evaluates the similarity where both the index terms and the stored texts are represented by weighted term vectors [Salton, 1991]. The assumption that index terms are independent (i. e., orthogonal vector space) obviously implies a short­coming of this retrieval method.

Within the domain of medicine, various KB systems have been developed and current systems will be briefly characterized here. Again, all projects in this area originate in the USA, with the exception of two (namely OSM and Medwise). One approach (namely CONSULTANT) addresses the field of veterinary medicine. A tabulated survey of the situation created in 1987 has been published [Kerkhof, 1987]. An occasional system is in the public domain, whereas others sell at a single-user price anywhere between $125 and $2000 depending on the configuration, discount, and other factors.

CMIT developed by the AMA [Finkel, 1981] forms a reference for the selection of preferred medical terms including certain synonyms, and generic terms with built-in arrangements to provide maximum convenience in usage, currency, and timely publication. Thus, CMIT is both a system of reference and a “distillate of a vast amount of medical knowledge.” The system is available in electronic form.

Blois was the first physician to apply CMIT as a diagnostic tool in his RECONSIDER project [Blois, 1984]. The application was released in 1981 and covered 3,262 disease entities, while 21,415 search terms were listed in a directory along with their frequency of occurrence (serving as an indicator of impact for each search term).

DXplain [Barnett et al., 1987] is also based on CMIT [Finkel, 1981] and was released in 1986. The project has close connections with AMA, and information is distributed using the World Wide Web. The KB contains information on 2,000 diseases and understands over 4,700 terms, with 65,000 disease-term relationships.

QMR patient diagnostic software [Middleton, 1991] covers 660 disease profiles and over 5,000 findings. It is the personal computer version of the INTERNIST-I prototype, sold by First DataBank in San Bruno at $495. The program is versatile, easy to use, and offers an attractive user-interface, but the size of its KB remained remarkably constant over the years.

The program MEDITEL [Waxman and Worley, 1990] addresses the issue of diagnosis in adults, and was marketed by Elsevier Publishers. Over the last few years, not much news has been reported in the literature.

ILIAD version 4.0 [Bergeron, 1991] is a software package designed to aid students and residents in their clinical decision logic. The ILIAD project is an outgrowth of the Health Evaluation through Logical Processes (HELP) system developed at LDS hospital in Salt Lake City, UT. The KB includes 1,300 diseases and 5,600 manifestations, and is built upon the experience of experts in nine subspecialties of internal medicine. The KB is updated semi-annually for revision and expansion. Future versions will support access to video laser discs, thereby providing an actual view of X-rays, echocardiograms, skin lesions, and other physical findings.

The Oxford System of Medicine (OSM) project was initiated by the Imperial Cancer Research Fund for use in primary care, and to help general practitioners during routine clinical work to support decision­making tasks. Such tasks include (1) diagnosing illnesses, (2) planning investigations, and patient treat­ment schedules, (3) prescribing drugs, (4) screening for disease, (5) assessing the risk of a particular disease, and (6) determining a referral to a specialist [Krause et al., 1993].

Medwise is a medical KB founded in 1983 and now covers some 3,900 disease entities, with 29,000 different keywords [Kerkhof et al., 1993]. It includes a separate KB with almost 500 equivalent terms, each referring to an average of three related terms e. g., extremity, leg, and limb. Equivalents are auto­matically generated to assist the user during the process of data entry. The matrix structure of the Medwise KB permits semantic differentiation of particular observations, e. g., lymphadenopathy as detected during palpation versus lymphadenopathy as interpreted from an X-ray picture. In the Medwise system both refer to different elements of the matrix, and their corresponding weight factors are individually incorporated when calculating the score for disease profile matching. For obvious reasons the language employed is English, but a complete dictionary that offers translations into Dutch is available [Kerkhof, 1996].

The Framemed system [Bishop and Ewing, 1993] divides medical information into 26 domains (com­parable to the matrix in Medwise) [Kerkhof et al., 1993], and arranges the items in a hierarchical sequence; formation of the hierarchies yields a logical framework for a standardized terminology that is in the public domain. The objective is to achieve a standard coded terminology to which all existing systems can relate, with obvious use as an electronic encyclopedia and for differential diagnosis. Framemed offers synonyms, and claims as an advantage a simple overall structure that facilitates updating.

STAT!-Ref [STAT, 1994] offers the contents of a first choice medical library (including several standard textbooks, e. g., on Primary Care) as well as Medline (either primary care, cardiology, or oncology) on CD-ROM.

MD-Challenger [MD-Challenger, 1994] offers a clinical reference and educational software for acute care and emergency medicine (everything from abdominal pain to zygoapophyseal joint arthritis), with nearly 4,000 annotated questions and literature references. The system also includes Continuing Medical Education (CME) credits.

Labsearch/286 is a differential diagnosis program allowing input of up to two abnormal laboratory findings plus information on symptoms and signs. Laboratory data concentrate on body fluids (blood, urine, cerebrospinal, ascitic, synovial, and pleural fluid) entered as high or low [Labsearch, 1994]. The system includes 6,500 diseases and 9,800 different findings. No details are available on its diagnostic performance, although the restriction of only two laboratory data certainly will limit its potential.

CONSULTANT is a KB for veterinary medicine developed by White [White and Lewkowicz, 1987]. This database for computer-assisted diagnosis and information management is available on a fee-for – service basis and is actually used in hundreds of private practices and institutions in North America.

Dambro [1998] compiled a book (approx. $50) that is updated annually, Griffith’s 5 Minute Clinical Consult. The first edition appeared in 1993 and indeed contains a realm of chart-like presented practical information on a thousand topics with reference to their ICD-code. A CD-ROM version is also on the Market. Similar to the design of Framemed [Bishop and Ewing, 1993], the information is compiled by a group of contributing authors whose names are listed in conjunction with each disease profile. Tilly and Smith [1997] presented a comparable book for the field of veterinary medicine along with a CD – ROM version: The 5 Minute Veterinary Consult.

The Birth Defects Encyclopedia edited by Buijse [1992] has the significant subtitle “The comprehensive, systematic, illustrative reference source for the diagnosis, delineation, etiology, biodynamics, occurrence, prevention and treatment of human anomalies of clinical relevance.” A unique feature of this printed encyclopedia is the BDFax-service which allows anyone in the world to call the Center for Birth Defects Information Services [Buyse, 1992] and request a current, daily updated version of any article in the KB. The information will then be faxed. Furthermore, BRS Information Technologies makes the full text available for online search and retrieval. Finally, the related Birth Defects Information System (BDIS) is a sophisticated computer-based profile matching system that reduces the research and diagnostic assis­tance tasks associated with complex syndromes.

Problems Related to Medical Terminology

Kiotomy (ki-ot’o-me) excision of the vulva (in Dorland’s pocket Medical Dictionary, 21st ed., p. 342; of vulvectomy on p. 684).

Inspection of other sources suggests that in the word vulva the first letter v should be replaced by the letter u, followed by rearrangement of letters to obtain an anagram, namely uvula (being a portion of the soft palate).

Medical language forms one of the greatest obstacles for the practical use of any type of KB designed for application in the field of medicine [Kerkhof, 1992]. Just imagine the confusion introduced when a visiting surgeon from an exotic country is requested to perform a kiotomy, and consults the above mentioned popular dictionary to look up what is supposed to be excised.

Natural language often has remote roots e. g., adrenaline and epinephrine are the same chemical sub­stances [Kerkhof, 1996]. Various words can be used in two senses: as noun and as adjective, e. g., antipyretic and adrenal. The Latin word “os” means both “mouth” and “bone”. And what’s in a phrase? “… a neurologist found a nonfluent dysphasia, without circumlocution or paraphasias” (Is there a single term for computer entry?). Or: “The CT scan showed accentuation of the peripheral margins of the bilateral parieto-occipital forceps major, and splenium low-absorptive abnormalities.” What does it mean, anyway? These examples illustrate the problem of translating medical phrases into concise “computer-storable language”.

Besides problems inherent in the understanding of natural language, additional difficulties pertaining to medical terminology can be indicated:

American vs. British spelling. Two standard differences are evident, namely the use of the digraph in British spelling (e. g., anaemia vs. anemia) and preference for using c (e. g., in leucocyte) rather than k (as in the American word leukocyte). However, combination of both rules does not apply in the British equivalent of the American spelling of the word leukemia, where the anglicized version is spelled as “leukaemia” [Kerkhof, 1996].

Synonyms. For some reason “epistaxis” is identical to “nosebleed”, and “pruritus” equals “itching”; it is not all that difficult, but the major problem is that you have to recall this every time you use either of them. Thrombocytosis and thrombocythemia are two words to indicate that the number of platelets in the peripheral circulation is in excess of 350,000 per microliter.

Eponyms. Many disease names refer to the first author (e. g., Boeck’s disease for sarcoidosis) who described the particular disorder, to the first patient analyzed in detail (e. g., Mortimer’s disease, again for sarcoidosis), or to the area (e. g., Lyme disease) where the illness was first detected. An epidemiologist ever attempted to replace the term Bornholm disease just to honor a friend (Sylvest). Geographical variations also occur: Grave’s disease (or Parry’s disease) as it is known in the U. S., U. K., and Australia (referring to goiter), is termed Basedow’s disease on the European continent, but Flajani’s disease in Italy.

Preferred terminology. In radiology “air” means gas within the body, regardless of its composition or site, but the term should be reserved for inspired atmospheric gas. With reference to pneumotho­rax, subcutaneous emphysema, or the contents of the gastrointestinal tract, the preferred term is “gas”. Sometimes the preferred terminology refers to simplicity; there is no virtue in talking about “male siblings” if we mean “brothers”. On other occasions the preferred terminology pertains to technical vocabulary which permits high precision and resolution descriptions if the available information is extremely exact. In these circumstances a valuable tool is blunted, if carelessness creeps in. For example the word “clumsiness” describes defective coordination of movement, whereas the term “dysdiadokokinesis” refers to the well-defined phenomenon of a defect in the ability to perform rapid movements of both hands in unison [Murphy, 1976].

Different expressions and their exact meaning. A straightforward example is: tympanism, tympanites, tympanitis and tympany, particularly in relation to: tympanal, tympanic, tympanous, and tym­panitic [Kerkhof, 1996]. One annotated example is presented: (1) heterogenous (not originating in the body); (2) heterogeneous (not of uniform composition, quality or structure); (3) hetero­genetic (pertaining to asexual generation); and (4) heterogenic (the same as heterogeneous).

Implicit information. A particular statement may imply a multitude of information components, e. g., if urinalysis is found to be normal, then the preceding examination implies (at least) the absence of the quadruplet proteinuria, hematuria, glucosuria, and casts. Also mirror-terms may apply: “dry cough” implies “no productive cough” (which can be entered as a negative finding). Likewise, leukopenia in particular implies “no leukocytosis”. This mutual exclusion principle applies to all antonyms, especially all terms beginning with hypo – or hyper-.

Imprecise terminology [Yu, 1983]. Some terms may carry a vague meaning, e. g., tumor, swelling, mass, and lump. To a large extent, however, the use of such terms reflects the uncertainty around an observation. In that respect it refers to a justifiable “law of preservation of uncertainty”. In other words, it would be incorrect to specify an observation in greater detail than the facts permit. This notion has consequences for the selection of equivalents.

Certainty vs. uncertainty. Decision analysis itself does not reduce the uncertainty about the true state of nature, but as long as some choice needs to be made it does enable one to make rational decisions in the light of uncertainty [Weinstein and Fineberg, 1980]. Yet another aspect of certainty vs. uncertainty deserves mentioning, namely where percent-wise figures about prognosis or outlook are subjectively interpreted [Eraker and Politser, 1982]. Reduction of the probabilities by a factor of one-tenth completely reversed the preferences of a set of respondents. This “certainty effect” shows that outcomes perceived with certainty are overweighted relative to uncertain outcomes. Thus, the formulation of information affects its interpretation by humans.

Knowledge engineering implies various levels of translation [Adeli, 1990]. The expert formulates as precisely as possible his thoughts, the engineer provides feedback using his/her own phrases to ensure an exact match between both minds, and subsequently the resulting expression is translated to a format usable for the computer program. These three steps involve transformations of language, while assuming that the occasional user of the program appreciates the full scope of the original thoughts of the expert.

Information source vs. actual patient [Kerkhof, 1987]. Current medical information sources tend to adhere to preference terminology to promote the use of uniform medical language. However, such standard vocabulary is not used by the average patient to describe their individual health problems [Hurst, 1971]. Then it is left to the clinician to transpose, e. g., “puffy face” and “moon face” if appropriate. Indeed, better health care can be realized by educating the patient about the value of structured communication with the physician [Verby and Verby, 1977].

Subspecialty interpretation and jargon. In chest radiology “type I bulla” means the same as “bleb” or “air cyst”, while pneumatocele is not regarded as a proper synonym [Glossary, 1989]. When naming a “hollow space” you may choose anything out of the following set: (1) cavity, (2) crypt, (3) pouch, (4) gap, (5) indentation, (6) dell, (7) burrow, (8) crater, (9) concavity, (10) excavation, (11) gorge,

Pocket, (13) cave, (14) cavern, (15) cistern, or (16) lacuna. However, every expression may exhibit a nuance within a certain context. Then there is a jargon: the terms (1) show, (2) engage­ment, (3) lightening, and (4) station, for example, have a particular meaning within the field of obstetrics [Kerkhof, 1996]. The term “streaking” has a meaning which differs for the microbiologist and the radiologist.

Multilingual approaches. The relation between a concept and the various corresponding terms in different languages is, in general, not unique. This implies that a multitude of different words from different syntactical categories may represent a single concept. Particularly the European countries are confronted with additional natural language problems. The Commission of the European Communities supports research activities in this area such as EPILEX (a multilingual lexicon of epidemiological terms containing Catalan, Dutch, English, French, German, Italian, Portuguese, and Spanish) [Du V Florey, 1993] and the development of a multilingual natural language system [Baud et al., 1994].

Frequency of occurrence. The meaning of semi-quantitative indicators like “always, often, etc.” is not transparent when screening a medical text. The intuitive interpretation of some quasi-numerical determinants is summarized elsewhere [Kong et al., 1986].

Noise terms. A study [Kerkhof, 1993] revealed that input information consisted on average of 75 terms per patient case; required for establishing the primary diagnosis were only 15 terms. This implies that 80% of the input data consisted of “noise terms”, which do not directly contribute to the confirmation of the correct diagnosis. Rather such overwhelming portions of information may blur the process of hypothesis formation, and in particular when the diagnostic task is solely left to the human being who is prone to confusion if the number of contributing elements is more than about ten.

Solution for Discrepancies

“Representation of knowledge is an art, not a science” [D. B. Lenat, 1989].

Various scientists have described techniques to help and solve language-related problems in the commu­nication area of medicine. An available online medical dictionary can be an important tool for research and application in natural language processing. Dorland’s Illustrated Medical Dictionary has been con­verted to an online interactive computer-based version. To assure exact description of definitions, a coding system has been advocated because of: (1) ease of handling, (2) redundancy to avoid errors, (3) specificity,

Indication of relationships (e. g., by using common code groups for similar terms), (5) equating equivalent terms and linking them in different languages, and (6) the feasibility to cross-link terms on the basis of common portions of their codes [Bishop and Dombrowski, 1990].

Attempts have also been made to automatically translate existing medical terminology systems [Cimino and Barnett, 1990]. For example, using a semantic network for mapping, the closest match between the MeSH term “portography” was “portal contrast phlebogram” in the ICD-procedures directory.

Two independent routes are available to attack the majority of linguistic problems related to a medical KB: (1) Use uniform input data, i. e., force the user of the KB to enter only controlled terms as they already exist in the KB (terms allowed by the editorial committee responsible for the KB). This route implies tedious work for the user, and the job of selecting recognizable entries has to be carried out every time the KB is consulted. (2) Apply an additional KB with equivalent expressions. Once the user enters a particular term, the KB automatically generates a list of equivalent expressions to choose from. The construction of such an auxiliary KB has to be realized only once, apart from the obvious updating process desired as the system evolves. Furthermore, a matrix structure as applied in the Medwise KB proves to be helpful in organizing information, and incorporating a substantial portion of semantic interpretation when dealing with medical knowledge. Inclusion of the auxiliary KB also enhances user – friendliness, and accelerates the process of data entry [Kerkhof et al., 1993].


While advances in computer technology permit virtually unlimited storage and fast retrieval capabilities, interpretation of natural language still constitutes a major obstacle for the application of medical KBs. Even common words often have highly ambiguous meanings [Murphy, 1976], e. g., “weakness” can mean lethargy or paralysis. Therefore, the user of any KB should be aware of the fact that justifiable adoption of a particular equivalent expression may depend on the precise interpretation of the complete case report.

On the other hand, the use of uniform input data on the basis of a thesaurus would imply a substantial loss of refinement with respect to terminology. Ideally, there is a maximum degree of freedom in the selection of words and their nuances, both for the KB and for the description of the individual case study.

An auxiliary KB with synonyms [SNOMED, 1993; Gabrieli, 1992; Chisholm, 1990; Kerkhof et al., 1993; Lindberg et al., 1993] indeed may offer considerable help during the process of matching input terms with information sources [Kerkhof, 1993]. An adequate solution to (at least a part of the) linguistic problems, will substantially enhance the performance of diagnostic computer programs.


A multitude of projects address the issue of how to handle an ever increasing amount of medical information. Coding schemes have been designed and are regularly refined, while other approaches aim to collect, structure, and disclose this gigantic corpus of information that features an unconfined potential to thrive.

The lack of standardized medical language limits the optimal use of computers in medicine. Major obstacles concern: (1) imprecise terminology, (2) the limited scope of a thesaurus, (3) multilingual conversion problems, (4) discrepancies between information sources vs. actual clinical cases, as well as

The ever expanding range of synonyms, acronyms, antonyms and eponyms. Incorporation of a separate knowledge base containing equivalent expressions appears to be indicated for the practical use of medical information systems.

Defining Terms

AMA: American Medical Association, with headquarters in Chicago. This organization is actively

Involved in new developments concerning medical informatics. CMIT [Finkel, 1981] and CPT [Physicians’ Current Procedural Terminology, 1992] as well as the cooperation with DXplain [Barnett et al., 1987] are prototypes of their activities.

ICD: International Classification of Diseases, a widely accepted system to organize all possible medical

Diagnoses. The tenth version is translated for worldwide application.

IMIA: International Medical Informatics Association, consisting of the national societies from about

40 countries.

NLM: National Library of Medicine, in Bethesda (Maryland), USA.

UMLS: Unified Medical Language System, a project initiated by the NLM and distributed on CD-ROM.

Cooperation with parties to implement the system within their own environment is encouraged, but requires a contract [Lindberg et al., 1993].


Adeli, H. 1990. Knowledge Engineering, Vol. I and II. New York, McGraw-Hill.

Barnett, G. O., Cimino, J. J., Hupp, J. A., and Hoffer, E. P. 1987. DXplain, an evolving diagnostic decision – support system. JAMA 258:67-74.

Baud, R., Lovis, C., Alpay, L., Rassinoux, A. M., Scherrer, J-R., Nolan, A., and Rector, A. 1994. Modelling for natural language understanding. SCAMC 93:289-293.

Baxt, W. G. 1991. Use of an artificial neural network for the diagnosis of myocardial infarction. Ann. Intern. Med. 115:483-848.

Bergeron, B. 1991. ILIAD, a diagnostic consultant and patient simulator. M. D. Comput. 8:46- 53.

Bishop, C. W. and Dombrowski, T. 1990. Coding, why and how. M. D. Comput. 7:210-215.

Bishop, C. W. and Ewing, P. D. 1993. Framed, a prototypical medical knowledge base of unusual design. M. D. Comput. 10:184-192.

Blois, M. S. 1984. Information and Medicine. Berkeley, Univ. California Press.

Buyse, M. L. (Ed.), Birth Defects Encyclopedia. Center for Birth Defects Information Services, Inc., Dover, MA; BDFax service (508) 785-2347.

Chisholm, J. 1990. Read Clinical Classification. BMJ 300:1092.

Cimino, J. J. and Barnett, G. O. 1990. Automated translation between medical terminologies using semantic definitions., M. D. Comput. 7:104-109.

Dambro, M. R. 1998. Griffith’s 5 Minute Clinical Consult. Baltimore, Williams & Wilkins.

Dorland’s pocket MEDICAL DICTIONARY, 21st edition p.342; cf. vulvectomy on p.684.

Du V Florey, C. 1993. EPILEX, a multilingual lexicon of epidemiological terms, Office for official publications of the European Communities (ISBN 92-826-6211-X), Luxembourg.

Eraker, S. A. and Politser, P. E. 1982. How decisions are reached; physician and patient. Ann. Intern. Med. 97:262-268.

Finkel, A. J., Ed. 1981. Current Medical Information and Terminology, 5th ed. American Medical Associa­tion, Chicago.

Gabrieli, E. R. 1992. A New American Electronic Medical Nomenclature, Automedica 14:23-30.

Glossary of “chest” radiological terms. 1989. Nomenclature Committee of the Fleischner Society.

Hurst, J. W. 1971. The art and science of presenting a patient’s problems. Arch. Intern. Med. 128:463.

International Classification of Diseases, tenth Clinical Modification. 1994. Commission on Professional and Hospital Acitivities, Ann Arbor, Michigan.

Kerkhof, P. L.M. 1987. Dreams and realities of computer-assisted diagnosis systems in medicine (Edito­rial). Automedica 8:123-134.

Kerkhof, P. L.M. 1992. Knowledge base systems and medical terminology. Automedica 14:47- 54.

Kerkhof, P. L.M. 1996. Woordenboek der geneeskunde N/E and E/N. (Medical Dictionary, 2nd ed.), Houten, The Netherlands, Bohn Stafleu Van Loghum (ISBN 90-6016-858-5).

Kerkhof, P. L.M. 1998. Veterinair Woordenboek E/N and N/E. (Veterinary Dictionary), Maarssen, The Netherlands, Elsevier Publ. (ISBN 90-6348-240-X).

Kerkhof, P. L.M., van Dieijen-Visser, M. P., Koenen, R., Schreuder, J. J.A., de Bruin H. G., and Gill, K. 1993. The Medwise Diagnostic Module as a Consultant. Expert Systems with Applications 6:433-440.

Kong, A., Barnett, G. O., Mosteller, F., and Youtz, C. 1986. How medical professionals evaluate expressions of probability. New Engl. J. Med. 315:740-744.

Krause, P., Fox, J., O’Neil, M., and Glowinski, A. 1993. Can we formally specify a medical decision support system? IEEE Expert 8:56-61.

Labsearch/286, software available from REMIND/1 Corp., P. O. Box 752, Nashua, NH 03061, USA.

Lenat, D. B. and Guha, R. V. 1989. Building Large Knowledge Base Systems. Addison-Wesley Publ. Co.

Lindberg, D. A., Humphreys, B. L., and McCray, A. T. 1993. The Unified Medical Language System. Methods Inf. Med. 32:281-291.

MD-Challenger, 70 Belle Meade Cove, Eads, TN 38028, USA.

Medical Records Institute, P. O. Box 289, Newton, MA 02160, USA.

MeSH, Medical Subject Headings. 1993. National Library of Medicine, Bethesda.

Middleton, B., Shwe M. A., Heckerman D. E., Henrion M., Horvitz E. J., Lehmann H. P., and Cooper G. F. 1991. Probabilistic diagnosis using a reformulation of the INTERNIST-1/QMR knowledge base; part II evaluation of diagnostic performance. Meth. Inform. Med. 30:15-22.

Murphy, E. A. 1976. The Logic of Medicine. Johns Hopkins Univ. Press, Baltimore, MD.

Physicians’ Current Procedural Terminology 4th ed., 1992. American Medical Association, Chicago.

Rouechй, B. 1982. The Medical Detectives, Washington Square Press, New York, p. 317.

Salton, G. 1991. Developments in automatic text retrieval. Science 253:974-980.

Smith, R. P. and Edwards, M. J.A. The Internet for Physicians. New York, Springer (ISBN 0- 387-94936-4).

SNOMED International. 1993. College of American Pathologists, Northfield, IL.

Sowa, J. F. 1984. Conceptual Graphs, Information Processing in Mind and Machine. Reading, MA, Addison – Wesley.

STAT!-Ref. 1994. Available from Teton Data Systems, P. O. Box 3082, Jackson, WY.

Tilly L. P., and Smith W. K. 1997. The 5 Minute Veterinary Consult. Baltimore, Williams, and Wilkins. Verby, J. and Verby, J. 1977. How to Talk to Doctors. Arco Publ. Co., Inc., New York.

Waxman, H. S. and Worley, W. E. 1990. Computer-assisted adult medical diagnosis, subject review and evaluation of a new microcomputer-based system. Medicine (Baltimore) 69:125-136.

Weed, L. L. 1991. Knowledge Coupling. New York, Springer-Verlag.

Weinstein, M. C. and Fineberg, H. V. 1980. Clinical Decision Analysis. W. B. Saunders, Philadelphia p.61. White, M. E. and Lewkowicz, J. 1987. The CONSULTANT database for computer-assisted diagnosis and information management in veterinary medicine. Automedica 8:135-140.

Williams, B. T. 1982. Computer Aids to Clinical Decisions. CRC Press Inc., Boca Raton, Florida Vol. II pp. 81-99.

Wyatt, J. 1991. Computer-Based Knowledge Systems. Lancet 338:1431-1436.

Yu, V. L. 1983. Conceptual Obstacles in Computerized Medical Diagnosis. J. Med. Philos. 8:67-83.

Further Information

The fields of medical knowledge bases and medical terminology are rapidly developing. There is no single comprehensive source of information available, but the reader is advised to scrutinize the following journals for new information:

M. D. Computing, published by Springer Verlag (New York and Berlin), reports on research in the field of medical informatics. Of special interest to the clinician.

IEEE Expert appears four times a year and presents the latest on artificial intelligence and expert systems. Contact P. O. Box 3014, Los Alamitos, CA 90720. For those interested in technical details on intelligent systems and their applications. IEEE (P. O. Box 1331, Piscataway NJ 08855-1331, USA) pub­lishes a number of related journals e. g., on knowledge engineering and on multimedia.

The National Library of Medicine (NLM, 8600 Rockville Pike, Bethesda MD 20894, USA) releases news bulletins, and provides information on UMLS and contracts for cooperation.

Ongoing projects on medical nomenclature are surveyed elsewhere in this chapter.

Pertinent Internet Web Sites:

British Medical Journal: Http://www. bmj. com JAMA: Http://www. ama-assn. org/public/journals The Lancet: Http://www. thelancet. com

The New England Journal of Medicine: Http://www. nejm. org Grateful Med: Http://igm. nlm. nih. gov/

Physician’s Desk Reference: Http://www. pdrnet. com/

About ICD-9-CM: http:YWww. icd-9-cm. org/

About ICD-10: http:Www. cdc. gov. nchswww/about/major/dvs/icd10des. htm

Medical guide: Http://www. medmatrix. org

Search engine for Health on the Net: Http://www. hon. ch/

Alternative Medicine: Http://www. yahoo. com/health/alternative_medicine Dermatology images: Http://tray. dermatology. uiowa. edu/dermimag. htm Focus on molecular medicine: Http://www. aston. it/icamm

Obviously, annual meetings form the forum for presentation of the latest developments: IMIA con­ferences are well known besides the world congress organized every four years. IMIA publishes a Yearbook of Medical Informatics; contact Schattauer Publishers, Subscriptions Dept., P. O. Box 104545, W-7000 Stuttgart 10, Germany.

Johnson, S. B. “Natural Language Processing in Biomedicine.” The Biomedical Engineering Handbook: Second Edition.

Ed. Joseph D. Bronzino

Boca Raton: CRC Press LLC, 2000


In the earliest years of computers in the 1930-40s, computers were sometimes dedicated to a single task, but were far too large and expensive for most kinds of tasks performed by embedded computers of today. Over time however, the concept of programmable controllers evolved from traditional electromechanical sequencers, via solid state devices, to the use of computer technology.

One of the first recognizably modern embedded systems was the Apollo Guidance Computer, developed by Charles Stark Draper at the MIT Instrumentation Laboratory. At the project’s inception, the Apollo guidance computer was considered the riskiest item in the Apollo project as it employed the then newly developed monolithic integrated circuits to reduce the size and weight. An early mass-produced embedded system was the Autonetics D-17 guidance computer for the Minuteman missile, released in 1961. It was built from transistor logic and had a hard disk for main memory. When the Minuteman II went into production in 1966, the D-17 was replaced with a new computer that was the first high-volume use of integrated circuits. This program alone reduced prices on quad nand gate ICs from $1000/each to $3/each, permitting their use in commercial products.

Since these early applications in the 1960s, embedded systems have come down in price and there has been a dramatic rise in processing power and functionality. The first microprocessor for example, the Intel 4004, was designed for calculators and other small systems but still required many external memory and support chips. In 1978 National Engineering Manufacturers Association released a "standard" for programmable

Microcontrollers, including almost any computer-based controllers, such as single board computers, numerical, and event-based controllers.

As the cost of microprocessors and microcontrollers fell it became feasible to replace expensive knob-based analog components such as potentiometers and variable capacitors with up/down buttons or knobs read out by a microprocessor even in some consumer products. By the mid-1980s, most of the common previously external system components had been integrated into the same chip as the processor and this modern form of the microcontroller allowed an even more widespread use, which by the end of the decade were the norm rather than the exception for almost all electronics devices.

The integration of microcontrollers has further increased the applications for which embedded systems are used into areas where traditionally a computer would not have been considered. A general purpose and comparatively low-cost microcontroller may often be programmed to fulfill the same role as a large number of separate components. Although in this context an embedded system is usually more complex than a traditional solution, most of the complexity is contained within the microcontroller itself. Very few additional components may be needed and most of the design effort is in the software. The intangible nature of software makes it much easier to prototype and test new revisions compared with the design and construction of a new circuit not using an embedded processor.