PII: S0957-4174(94)E0001-B Pergamon Expert Systems With Applications, Vol. 8, No. 1, pp. 89-99, 1995 Copyright © 1994 Elsevier Science Ltd Printed in the USA. All rights reserved 0957-4174/95 $9.50 + .00 0957-4174(94)E0001-B An Expert System for Homeopathic Glaucoma Treatment (SEHO) F. ALONSO-AMO, A. G r M E Z PI~REZ, G . L O P E Z G O M E Z , A N D C. M O N T E S Facultad de lnform~itica--Universidad Politrcnica de Madrid, Campus de Montegancedo--Boadilla del Monte, Madrid, Spain Abstract--In this article, an Expert S y s t e m f o r Homeopathic Glaucoma Treatment ( S E H O ) is pre- sented, the task o f which is to assist ophthalmologists in selecting the most appropriate therapy f o r a patient diagnosed as having glaucoma. It is based on techniques proper to homeopathic medicine, a trend that is gaining more and more supporters all over the world, but in which real experts are f e w and f a r between. After a brief overview o f the state o f the art, the authors describe in detail on the development o f the system, f o r which the I D E A L methodology, designed f o r knowledge-based system development, was used. 1. I N T R O D U C T I O N LAMENTABLY, GLAUCOMA is o n e o f the most signifi- cant pathologies affecting sight, the n u m b e r o f cases making it one o f the m a j o r visual disorders. This disease often leads to irreversible blindness and is accompanied by an inherited t r a u m a t i c process, m a k i n g g l a u c o m a sufferers especially sensitive to the m e t h o d s used in their t r e a t m e n t . A m a r k e d partiality toward the field o f alternative m e d i c i n e has been observed within the blind com- m u n i t y , o n some occasions as a c o m p l e m e n t to tra- ditional m e d i c i n e a n d o n others due to their openness to a n d acceptance o f these branches o f medicine, o f which h o m e o p a t h y is a clear example, which they con- sider to be less aggressive. H o m e o p a t h i c m e d i c i n e is based o n the activation o f the organism's healing mechanisms by administering h o m e o p a t h i c dilutions c o r r e s p o n d i n g to specific doses o f these medicines. It should n o t be forgotten that a blind person faces o t h e r p r o b l e m s besides sight i m p a i r m e n t , such as dif- ficulties in social integration, adaptation, and mobility, etc., as well as other underlying illnesses which in m a n y cases originally i n d u c e d the process leading to blind- This paper was sponsored and backed by the ONCE (Spanish National Organization of the Blind) and the CICYT (Interministerial Com- mission of Science and Technology) grant no. TIC1235/12-E and prepared and written with the collaboration of CETTICO (Centre o f Technology Transfer in Knowledge Engineering, Madrid, Spain). Requests for reprints should be sent to F. Alonso-Amo, Facultad de Inform~tica--Universidad Politrcnica de Madrid, Campus de Montegancedo--Boadilla del Monte, 28660 Madrid, Spain. 89 ness. As a result, these people can hardly be treated medically w i t h o u t their physical an d m en t al charac- teristics being t ak en into account. H o m e o p a t h y is in fact o ri en t ed in this direction, and, as far as visual disorders are concerned, h o m e o - pathic t r e a t m e n t depends o n four factors (Rubio, 1988): the threat to vision, the k i n d o f disorder a n d m a n n e r in which it evolves, the level o f individual sensitivity, a n d the subject's potential for response. A n d such t r e a t m e n t is c o m p o s e d o f • Rem ed i es prescribed for local s y m p t o m s , an d • Rem ed i es for the g l a u c o m a t o u s field (patient's phys- ical a n d m e n t a l constitution). T h e secretion o f a q u e o u s h u m o u r (an i m p o r t a n t factor in glaucoma) is closely linked with the sensitivity o f the neurovegetative system a n d that the stress o f ev ery d ay life a n d e m o t i o n s affect pressure in the eye- ball. Psychological tests have revealed t h at g l au co m a sufferers are anxious, susceptible, a n d meticulous a n d that the g l a u c o m a t o u s subject's hypersensitivity affects the t r e a t m e n t o f the patient. H o m e o p a t h i c t r e a t m e n t is very often associated with traditional therapies, often permitting a r e d u c t i o n in the dosage a n d an increase in the patient's tolerance to the latter. So, we can say t h at the different therapies, far f r o m being opposed to each other, are co m p l em en t ary , or as Blamentier states (Rubio, 1988), " i t is the patients an d n o t the therapies t h at differ." It is i m p o r t a n t to observe h o w sensitive the patient is to a t h e r a p y that stimulates his defence potential. However, it is n o t easy to appreciate this potential, as it d ep en d s o n m a n y factors, such as age, the degree o f sensitivity o f the lesions, a n d the patient's character- istics an d constitutional data. These m a k e it possible 90 F. A l o n s o - A m o et al. for the homeopathic doctrine and clinical expertise to decide what is the most suitable treatment. However, it is not possible to establish a general rule to cover the wide range of possibilities that crop up in the homeopathic doctor's surgery and, at all events, the only suitable guide he or she has personal experi- ence. Therefore, it was thought necessary to employ knowledge engineering techniques to deal with this problem. This led to the development of the expert system described, which was developed at CETTICO (Centre of Technology Transfer in Knowledge Engi- neering) and is able to assign homeopathic treatment to glaucoma sufferers. There are no similar expert systems on record. Vi- sual disorders have been little dealt with in the past, and the therapy model has always been based on al- lopathic medicine. The CASNET system (Kulikowski & Weiss, 1982), oriented to the diagnosis and treatment of different kinds of glaucoma, is an example of this. In addition, there is a knowledge gap in homeopathic glaucoma treatment, due both to the lack of experts in this kind of therapy and to the fact that they are con- centrated in half a dozen countries. Therefore, the system entails a qualitative advance in the treatment of the blind, bringing innovative tech- niques into an alternative approach to medicine that is held in high esteem by the blind community. 2. C O M P U T E R SYSTEMS IN H O M E O P A T H I C MEDICINE Until recently, the computer models used to simulate medical decision making in a computer system have been mainly based on probabilistic methods, as their machine representation is easy to obtain (Gorry, Sil- verman, & Pauker, 1978; Solomon & Papert, 1976). Despite the fact that these programs have come up with very interesting results, the doctor does not iden- tify his or her reasoning and manner of arriving at a diagnosis with theirs, and it is also difficult to evaluate the quality of a diagnosis proposed in this way. In addition, it has been noted (Ledley & Lusted, 1979) that the majority of clinical errors are made by omission, that is, errors due to a failure to take into account all of the possibilities playing an important role in determining the illness suffered by the patient so as to arrive at the correct diagnosis and treatment. Therefore, a doctor needs assistance in establishing the diagnosis and a suitable therapy, especially in the case of unusual illnesses or when the patient's symptoms may lead to different interpretations. Considering that all of the information required on a patient can be stored and classified in a computer, along with the symptoms of the illnesses of a domain, it follows that, in such circumstances, a computer may come up with a more precise and rapid response than a doctor (Barr & Feigenbaum, 1982), especially when the knowledge of the symptoms of an illness has been elicited from an expert doctor and incorporated into a knowledge base that interacts with an expert system. A complete and adequate homeopathic study of a patient depends on the skill of the homeopath and his or her ability in identifying, storing, recording, refer- encing, analysing and evaluating any class of data or group of data. This requires a system of classification, according to which the concepts and relevant infor- mation are organized, and a coding, which facilitates their use. Originally, the homeopath's traditional prescription and, later, data bases, which were and are of great help in homeopathic surgeries, were used for classification and coding. However, it has been noted on several oc- casions that doctors are generally somewhat reluctant to use computers as a tool and consider them to be little suited to establishing repertories, that is, what medicines cover the patient's symptoms, the number of symptoms, and to what extent. On the other hand, medical reasoning is related to judgement problems, problem solving, decision mak- ing, and knowledge (Fieschi, 1987), which is why it has come to be a traditional working domain in knowledge engineering. The introduction of ES into medical diagnosis and treatment has done away with initial scepticism, and they have come into more widespread use. Examples of this are MYCIN (Shortliffe, 1976), TEIREISIAS (Davis, 1976), INTERNIST (Pople, 1977), PIP (Pauker & Szolovitz, 1977), DIGITALIS THERAPY ADVI- SOR (Gorry et al., 1978), CENTAURI (Aikins, 1980), SAM (Gascuel, 1981), ATTENDING (Miller, 1988), CASNET (Kulikowski & Weiss, 1982), NESTOR (Cooper, 1984), KARDIO (Bratko, 1989). Unfortunately, however, few ES have gained access to homeopathic surgeries, though several computer systems based on this alternative medicine have been developed, such as the following. • RADAR (Shrogens, 1982), which contains several pharmacopoeias, including those by Allen, Hering, Heneman, and Boerick. It has access to 2,000 ho- meopathic remedies and their corresponding phar- macopoeias. • HINEIRO (Bachelerie, 1986) contains 2,535 Boen- ninghausen therapy rubrics (symptoms). Each of these rubrics is associated with a blackboard with the most common remedies. • ABIES (Benson, 1980) is a clinical information sys- tem. In addition to carrying out medical treatment, it locates notes and treatments for patients using the RCC system (Real Clinical Classification) (Read & Benson, 1986), which is a hierarchical statistical classification of a nomenclature with four detail levels. • STAPHISE (SalaiJn & Simonet, 1989), an infor- mation system using the Ken repertory, composed Homeopathic Glaucoma Treatment 91 of some 20,000 rubrics taken from different phar- macopoeias. Its information base may be custom- ized. As regards ES in homeopathic medicine, we should mention VES (Vithoulkas, 1988), developed by the homeopath George Vithoulkas. His working philoso- phy can be situated within the unitarian current of homeopathy. The VES system returns the best remedy with a given scoring and certainty factor, indicating to what extent any alternative remedies presented can be administered. VES is integrated into the RADAR sys- tem and takes advantage of its potential. It is a general medical application and is not specialized in any class of illness. 3. SEHO (ES FOR HOMEOPATHIC GLAUCOMA TREATMENT) Considering homeopathy as complementary to tradi- tional treatments of visual disorders causing blindness and taking into account its acceptance among the blind community, it was thought necessary to research and develop an ES to treat glaucoma. This would assist the homeopath in inference tasks and, finding the medi- cines most suited to each patient, would come up with the appropriate dilutions. The end result was the pro- totype system SEHO (Cristrbal & Ortiz Latierro, 1991). As opposed to other systems, NEOMYCIN for ex- ample, which mainly use the description of the patient's illness to select the therapy (Clancey, 1981; Clancey & Shortliffe, 1984), SEHO compiles extensive and com- plete information on the patient through homeopathic questioning before suggesting any remedy and thus does not point the doctor in any particular direction that might lead him or her to overlook important data on the patient during the session. The results of the questioning session are sent to the homeopathic techniques of the expert, which attempts to put together the most complete profile of the patient possible and to establish his or her characteristics and symptoms so as to apply the most suitable medicines. Optionally, the system can provide information on other possible, though less suited, medicines together with the patient's symptoms that each of them covers. Like other ES, SEHO provides information on its reasoning process, explaining the intermediate conclu- sions and why it selects and incorporates each medicine into the working memory. There are two different trends in homeopathic med- icine: the unitarian trend, which suggests only one medicine as a remedy (e.g., VES) and the pluralist trend, to which SEHO belongs, which proposes differ- ent dilutions of several medicines. So, SEHO's dilutions contain several medicines. SEHO is, therefore, the first ES for treating a visual disorder from the point of view of alternative medicine. Another important characteristic differentiating it from other systems is the fact that it has been designed to be used by the blind, available Braille adaptations hav- ing been incorporated into the prototype system. Its line of reasoning, based on the techniques of the chosen expert, the Chairman of the Association of Homeo- paths of Madrid, leads to scaled dilutions, that is, to the assignation of three or more medicines in most cases, some with low, some with intermediate, and others with high dilutions, except when any of these are superfluous. The inference process passes through several stages before arriving at these recommendations. The first stage, medicine determination, covers all of the pa- tient's symptoms, both those related with specific symptoms and their field characteristics. It selects the lowest possible number of the most suited medicines. Provisional dilutions are assigned in the second stage, and the scaled dilutions are established in the final one. SEHO was developed using GURU, and its knowl- edge base is composed of seven rule bases, which con- tain a total of 72 rules based on public and expert knowledge on glaucoma, and five data bases, contain- ing medicines and symptoms, categories and field. Its inference mechanism is backward chaining. The prototype has been designed in such a way that its knowledge base and data bases, containing phar- macopoeias including the latest findings with respect to homeopathic remedies for glaucoma, can be ex- panded. 4. SPECIFICATION AND DEFINITION OF SEHO The IDEAL methodology (Mat6 & Pazos, 1988), one of the most prestigious methodologies for ES devel- opment today, was used to define and develop the pro- totype system. This methodology brings together the most relevant ones in this area. Some of these meth- odologies are (Hayes-Roth, Waterman, & Lenat, 1983; Liebowitz & De Salvo, 1989; Waterman, 1986). Requirements definition was based on the following general concepts: • Interface faithfully reflecting the content and extent of homeopathic questioning on the patient's char- acteristics and constitutional data and his or her symptoms, and adapted for the blind. • Storage of complete pharmacopoeias on each med- icine. • Assignation of scaled dilutions of the remedies in the final treatment: low (4 CH), intermediate (7 CH), and high (15 CH) dilutions. • Assignation of the smallest possible number of nec- essary medicines. The adequacy test was carried out to evaluate the application, and the characteristics were grouped in four dimensions according to the IDEAL methodology: plausibility, justification, success, and adequacy, using 92 F. Alonso-Amo et aL Dimension TABLE 1 Evaluation of the Application Dimension Value (geometric mean) (Vci) M a x i m u m Dimension Value (geometric mean) (Vcmi) Plausibility 62 89.1 Justification 29.8 64.1 A d e q u a c y 48.4 61.9 Success 46.8 66.5 t = 4 Vg (General value) = ~ Vci/4 = 46.75. I=1 1=4 Vm (Maximum value) = ~ Vcmi/4 = 70.4. i=1 Vf (Final application value) = V ~ x 10 = 6.64 > 5. variable threshold values to accept or reject a charac- teristic a n d the geometric m e a n (Mat6, 1988; Pazos, 1989) to evaluate the task. T h e application was f o u n d to be suitable for t r e a t m e n t using an ES (threshold val- ues o f 6.64 > 5) (Table 1). S E H O is a decision s u p p o r t system for assigning h o m e o p a t h i c resources or remedies, which was con- sidered central for its definition. T h e r e is an essential difference in the m e t h o d used for assigning the appro- priate treatment: it does n o t directly take a c c o u n t o f the patient's pathology or type o f glaucoma, as allo- pathic medicine would; the disease's manifestation is truly decisive, that is, the patient's s y m p t o m s a n d con- dition are determined by his or her field characteristics. S E H O exclusively considers the s y m p t o m s typical o f glaucoma and the psychological profile and personal factors o f the subjects suffering from glaucoma. Surgical affections are excluded f r o m this framework, t h o u g h c o m p l e m e n t a r y t r e a t m e n t or t h e r a p y to i m p r o v e tol- erance to allopathic medicines m a y be assigned. Figure 1 shows the flow chart o f the S E H O prototype. T h e r e are two types o f knowledge i m p l e m e n t e d in the system: • Public knowledge, based o n h o m e o p a t h i c p h a r m a - copoeias (Barraza, 1980; L a t h o u d , 1988) in 5 data- bases, including lists o f medicines along with specific symptoms, the categories a n d the field, as well as a list o f antagonistic categories. • Expert knowledge, c o n t a i n e d in 7 p r o d u c t i o n rule bases, elicited f r o m the expert a n d based o n the op- erative a n d heuristic p r o c e d u r e used by the expert himself to prepare treatment. T h e rules will be fired using a backward chaining control strategy, until all the premises in its a n t e c e d e n t are true or false. • T h e response t i m e has to be short (less t h a n 5 min- utes). • During execution, the system offers i n f o r m a t i o n o n the medicines that are being considered to cover the s y m p t o m s , category, a n d field, as well as c o m m e n t s o n how it arrives at the scale o f dilutions. • Once executed, the system offers i n f o r m a t i o n on medicines in the final solution, the dilution in which the m ed i ci n e should be administered, dosage, length o f the t r e a t m e n t until the n ex t visit, s y m p t o m s a n d characteristics that the selected medicines cover, a n d medicines selected t h at are n o t included in the s o - l u t i o n along with a list o f the s y m p t o m s t h at they cover. An ideal solution is o n e that assigns a scaled dilution o r a single m ed i ci n e per dilution. This ideal situation does n o t necessarily have t o o c c u r in every case: this depends o n the patient a n d his o r h er characteristics. 5. C O N C E P T U A L I Z A T I O N A N D F O R M A L I Z A T I O N O F S E H O Knowledge acquisition for the conceptualization a n d subsequent formalization o f the knowledge base was based o n n o n s t r u c t u r e d interviews in the first stage. Cases were presented to the expert a n d the p ro t o co l was analysed d u ri n g this process in the second stage. In the final stage, very explicit structured interviews were a m e a n s o f solving the problems that arose. It was f o u n d t h at a h o m e o p a t h views a patient f r o m three different b u t c o m p l e m e n t a r y angles, w h en estab- lishing a therapy: • Specific o r local sy m p t o m s, related with the organ o r organs affected b y the illness (eyes, in the case o f glaucoma). S y m p t o m s such as congestive p h e n o m - ena, increase in ocular pressure, sight i m p a i r m e n t , alteration o f the optical nerve, etc. • Categories, as a m e a n s o f classifying the sy m p t o m s. T h e y specify the i m p r o v e m e n t s o r deterioration o f a s y m p t o m o r o f the patient in general. • Field, which establishes h o w the patient reacts to the illness an d which is characterized b y the patient's characteristics: physical, mental, an d constitutional data, such as anxious, susceptible, meticulous, for example. T h e result was the conceptual m o d el shown in Fig- ure 2 a n d the knowledge m a p shown in Figure 3. P at ie n t M ed ic in e Q u es ti o n in g o n P at ie n t S y m p to m s Q u es ti o n in g o n P at ie n t C at eg o ry Q u es ti o n in g o n P at ie n t F ie ld C at eg o ry L o o k f o r S u it ab le M ed ic in es P ro d u ce S o lu ti o n M ed ic in es S o lu ti o n F IG U R E 1 . S E H O f lo w c h ar t. M ed ic in e B as e D el et e A ss ig n P ro v is io n al D il u ti o n s M ed ic in es a n d P ro v is io n al D il u ti o n s A ss ig n S ca le d D il u ti o n s J 94 F. Alonso-Amo et al. PATIENT 1 1 1 I I sY T° S I I ATIENTI I I MEDICINE , , SYMPTOMS SYMPTOMS I I CATEGORY n ~ I PAT1ENT CATEGORY I I MEDICINE CATEGORY I FIELD I~ n I I n I PATIENT MEDICINES I FIGURE 2. Conceptual model. Knowledge formalization required that a distinction be m a d e between two stages in the expert process: medicine d e t e r m i n a t i o n a n d dilution assignment. 5.1. Selection of Medicine Before medicines are selected, the patient's s y m p t o m s are elicited t h r o u g h questioning o n individual symp- toms, establishing their i m p o r t a n c e or otherwise, a n d o n the patient's categories a n d fields. Considering t h e x , y , z SI(x) nSI(x) n M O D ( x ) n M O S A N T A ( x ) n SNI(x) n T O T S I N T ( x ) ST(x) A B following definitions: Medicines Set o f i m p o r t a n t s y m p t o m s cov- ered b y m ed i ci n e x N u m b e r o f i m p o r t a n t s y m p t o m s co v ered b y medicine x N u m b e r o f categories covered b y m ed i ci n e x N u m b e r o f antagonistic categories o f m ed i ci n e x N u m b e r o f u n i m p o r t a n t specific s y m p t o m s co v ered by medicine x To t al n u m b e r o f s y m p t o m s cov- ered b y x (specific + category + field) Set o f field s y m p t o m s covered by x Set o f medicines selected for spe- cific sy m p t o m s. It changes as rules are fired a n d some medicines are selected an d others excluded. At the beginning o f the process, the set covers an y i m p o r t a n t sy m p t o m . Set o f medicines selected for the field. Like A, it changes w h en rules are fired. It initially covers a n y o f the patient's characteristics or sy m p t o m s. I Patient Name A Name Category Name A Pharmacopoeial Category 1 Med. Cat. Name Category Name Medicine Name Field Name Field Present in Pharmacopoeia Name, Symptom Importance Name. Important Name. Unimportant Symptom Name " ' " ~ 1 Name Symptom Name Medicine Symptom Name Medicine Name I Medicine N. Dilution / ] I Med. F. Name I " j ~v~e~c?aernI~la me I FIGURE 3. Knowledge map. Homeopathic Glaucoma Treatment 95 • T h e p r o c e d u r e followed b y the e x p e r t for selecting the m e d i c i n e to c o v e r given s y m p t o m s is based o n the following rules: I f x, y ~ A a n d SI(x) C SI (y), t h e n select y I f x, y ~ A a n d SI(x) = SI (y) a n d n M O D A N T A ( x ) > n M O D A N T A ( y ) , t h e n select y I f x, y E A a n d SI(x) = SI (y) a n d n M O D ( x ) > n M O D ( y ) , t h e n select y I f x, y ~ A a n d SI(x) = SI(y) a n d nSNI(x) > nSNI(y), t h e n select x I f x, y ~ A a n d SI(x) = SI (y) a n d nST(x) > nST(y), t h e n select x. T h e o b t e n t i o n o f the m i n i m u m set in A is based o n the rules below: I f x, y E A a n d [x,y] covers the s y m p t o m s o f Ix, y, z], t h e n A = A - [z]. I f there is m o r e t h a n o n e m i n i m u m set, select the o n e whose m e d i c i n e s c o v e r m o r e , i m p o r t a n t s y m p - t o m s . • T h e selection o f m e d i c i n e s for the field is centered o n the following rules: I f x, y E B a n d ST(x) C ST(y), t h e n select y I f x, y @ B a n d ST(x) = S T (y) a n d n T O T S I N T ( x ) > n T O T S I N T ( y ) , t h e n select x. T h e o b t e n t i o n o f the m i n i m u m set in B is b a s e d o n the s a m e rules as for the o b t e n t i o n o f the m i n i m u m A, t h a t is: I f x, y, z E B a n d [x, y] covers the s y m p t o m s o f [x, y, z], t h e n B = B - [z]. Select the m i n i m u m set t h a t covers m o r e , i m p o r t a n t s y m p t o m s . 5 . 2 . A s s i g n a t i o n o f D i l u t i o n s T h e goal p u r s u e d in this stage is the a s s i g n m e n t o f scaled dilutions. First a p r o v i s i o n a l a s s i g n m e n t o f the dilutions o f the m e d i c i n e s o b t a i n e d is taken, a n d the case-related o p t i m u m is sought. If: M H Set o f high-dilution m e d i c i n e s M L n H nL ML1 M L 2 nLi M H 1 M H i nHi Set o f low-dilution m e d i c i n e s N u m b e r o f m e d i c i n e s in M H N u m b e r o f m e d i c i n e s in M L Set o f m e d i c i n e s in M L t h a t c o v e r categories Set o f M L I m e d i c i n e s t h a t c o v e r m o s t cate- gories N u m b e r o f M L i m e d i c i n e s with i = l, 2. Set o f M H medicines that c o v e r field s y m p t o m s Set o f M H l m e d i c i n e s t h a t c o v e r m o s t s y m p - t o m s (i -- 2 . • . 9 ) N u m b e r o f M H i m e d i c i n e s with i = l • • • 9. T h e provisional dilutions are assigned according to the following rules: I f x E A a n d x E B, assign a low dilution to x a n d x @ M L I f x E A, assign a low dilution to x a n d x E M H I f A :~ 0 a n d B 4 = 0, t h e n provisional dilutions -- low I f B :~ 0 a n d M L v ~ 0, t h e n provisional dilutions = low a n d high I f A = B a n d B ~ 0 a n d A ~ 0, t h e n provisional dilutions = high T h e different p a t h s t a k e n t o arrive at the goal (the suit- able dilution), once the provisional dilutions have b e e n obtained, are represented in the shape o f a tree in Fig- ures 4, 5, a n d 6. T h e final t r e a t m e n t is b a s e d o n the following rules: I f x is assigned a low dilution, prescribe 3 doses o f a 4 C H dilution o f x p e r day. I f x is assigned a n i n t e r m e d i a t e dilution, prescribe 4 doses o f a 7 C H dilution o f x p e r day. I f x is assigned a high dilution, prescribe 5 doses o f a 15CH dilution o f x p e r day. 6. S E H O S Y S T E M I M P L E M E N T A T I O N T h e system p r o t o t y p e has b e e n i m p l e m e n t e d using G U R U as a d e v e l o p m e n t tool. In addition, a n interface, i n c o r p o r a t i n g a voice synthesizer a n d Braille line, has b e e n designed t o e n a b l e the blind to use the system. PROVISIONAL DILUTIONS = LOW [ nL = 1 nL>l / nL1 >_ 1 [ / nL> 1 I R14 Intermed, Dil, ] nL2>l J RI5 ] [Intermed. Dil. I x . , I nL1 = 1 ] R16 R17 FIGURE 4. Dilution assignation I. i INoc"an e'iodi' I bchaogesindi' I 96 F. Alonso-Amo et aL PROV. DILUTIONS = HIGH I FIGURE 5. Dilution assignation II. Its design and implementation matches the Struc- tures Map in Figure 7, in which the modules describe the following actions: MSINS Generates the interface for recording patients' symptoms MMODS Generates the category interfaces MTERRS Generates the field interfaces MLEESIN Searches the data base for medicines that cover any patient symptom MLEEMOD Searches the data base for medicines that cover any patient category MLEETERR ORDENA MINIMOSA ORDENB MINIMOSB MBUSDILU MDILBYA Searches the data base for medicines that cover any patient field Obtains final medicines for particular symptoms Obtains minimum sets of medicines for symptoms Obtains final medicines for field Obtains minimum sets of medicines for field Defines provisional dilution Defines suitable dilution I PROV. DILUTIONS = HIGH & LOW I / FIGURE 6. Dilution assignation III. "N l M A IN M O D U L E se h o D A T A C O L L E C T IO N M/ II M M~ O~ i I M TER RS C H O IC E O F S O L U T IO N M E D IC IN E S S E A R C H II F IG U R E 7 . S tr u ct u re s m a p . 98 M R E S U L T Assigns the most appropriate treat- m e n t , including medicines, their di- lutions, a n d the way o f administering them. 7. E V A L U A T I O N S E H O was evaluated in three phases in line with tra- ditional methodological orientations. 7.1. Validation of System Decisions by the E x p e r t Fifteen test cases that had been selected by the expert a n d set o u t in the project success criteria, along with a n o t h e r 20 o f the m o s t frequent cases p u t forward by the expert were used for this purpose. O f the examples selected, 10% were e x t r e m e cases a n d generated arti- ficially, 10% were ambiguous, a n d the remaining were typical cases. T h e expert a p p r o v e d the system's pro- cedure in all o f the cases. As regards dilution assignation, the different mod- ules were verified as follows. TABLE 2 Typical Case, C a s e No. 6 SS (Specific Symptoms) Nebulas Photophobia Orbital pain Reduced vision Dilated pupils Stigmata on the cornea Categories Worsens when lying down Worsens in the morning Worsens with changes in the weather Improves in the open air Improves with warmth FT (field symptoms or characteristics) Depression Obesity Apathy Pessimism Shyness Skin irritations Fatigue Constipation Egotism Recommended medicines FLUORIC CALCAREA--Iow dilution SULPHUR--intermediate dilution CARBONIC CALCAREA--high dilution CAUSTICUM--high dilution Optimal treatment with GELSEMIUM--only high NUX VOMICA--only high AURUM METALICUM~only high COMOCLADIA~only high Important (I) Unimportant (U) I I I I F. A l o n s o - A m o et al. A) Step f r o m provisionally low dilutions to low an d i n t e r m e d i a t e dilutions B) Step f r o m provisionally high dilutions to low a n d high dilutions C) Step f r o m provisionally low an d high dilutions t o low, intermediate, a n d high dilutions. It was f o u n d that, as required b y the expert, the dilu- tions are n o t fully scaled in two situations: • W h e n there are n o field characteristics and, therefore, low dilutions are n o t assigned; • It is impossible to assign intermediate dilutions o n the basis o f high a n d low dilutions, as the set o f med- icines c o u n t e d in this case is equal to the total. This case is considered extreme. Finally, it was f o u n d that the n u m b e r o f medicines assigned with a given dilution is n ev er greater t h a n 2, j u st as the expert stipulated. A typical case is illustrated in Table 2, indicating the f o r m t h ey take. 7.2. Validation of Typical Cases by Experts Not Involved With System Development Fifteen typical cases were p u t to t h em , an d they o n l y disagreed o n o n e a m b i g u o u s case. This was d u e to the fact t h at there were two equally acceptable forms o f t r e a t m e n t , a n d this was therefore a question o f pref- erences. Moreover, the system had indicated the second possibility as an optional treatment. T h e system has n o w b een transferred t o the Spanish N at i o n al Organization for the Blind (ONCE) as an aid for therapists n o t specialized in h o m eo p at h i c medicine. 8. FUTURE RESEARCH WORK Although the results provided b y S E H O are an im- p o r t a n t ad v an ce in the a u t o m a t e d t r e a t m e n t o f glau- c o m a using h o m e o p a t h i c techniques, we should n o t o v e r l o o k the fact t h at S E H O is a p r o t o t y p e requiring f u r t h e r d e v e l o p m e n t . Th i s will involve two courses o f action: o n e regard- ing the system's knowledge a n d the other, its c o m p u t e r structure. As regards the knowledge at present incor- p o r a t e d i n t o SEHO, it is p l a n n e d to ex t en d the medi- cine an d expert knowledge data bases. F o r this purpose, a n o t h e r expert in h o m e o p a t h i c medicine, likewise a m e m b e r o f the pluralist school, will j o i n the research t eam , with a view to adding to the knowledge a n d co m p ari n g his approach with that o f the f o r m e r expert. With respect t o the S E H O ' s c o m p u t e r structure, t h ree basic measures are envisaged. • T h e new knowledge acquisition stage will m a k e it possible to identify new rules an d new fo rm s o f pro- cessing an d handling the m o st useful aspects o f the individual cases. F o r this stage, it is p l a n n e d to in- c o r p o r a t e an a u t o m a t e d knowledge acquisition m o d u l e that will equip the system with the capability H o m e o p a t h i c G l a u c o m a T r e a t m e n t 99 of analyzing any new information that comes to light, as well as facilitating the knowledge acquisition task. There are also plans to equip the final system with an intelligent interface capable o f selecting the ques- tions to be put to the expert. • Adapt the user interface to a more flexible graphic environment for use by sighted personnel, while the present interface will be kept for the blind user. For this purpose, the screens o f yes/no questions (62 on symptoms, 57 on category, and 66 on field charac- teristics) will be replaced by 4 multiple choice win- dows with a scroll bar (for symptoms, category, and field, and to select the important symptoms from those chosen). The information will be output to an independent window controlled by the application, and new phases and explanations will be incorporated to make it easier to follow the system's logic. • With a view to creating an easily extendible and reusable expert system, in place of the present GURU-based configuration that is difficult to alter, the SEHO prototype will be adapted to an object- oriented environment by transforming the present medicine and knowledge bases and rules into a class structure. The CLASER (Set o f Open Libraries for the Development and Query of Reusable Expert Systems) (Garcia, 1993), recently developed at CET- TICO in C + + , will be employed to this end. From experience with other systems, this measure should make the future SEHO system approximately 10 times faster than the present system, in addition to the above-mentioned advantages o f extendibility and reusability. REFERENCES Aikins, J.S. (1980). Prototypes and Production Rules: A Knowledge Representation for Computer Consultations. Stanford Heuristic Project, Department of Computer Science Report No. STAN- C5-80-84. Bachelerie, R. (1986). HOMEOREP. Clermont-Fen'and. Barr, A., & Feigenbaum, E.A. (1982). The handbook of artificial in- telligence. Reading, MA: Addison Wesley. Barraza, J. (1980). Elementos de terapedaica homeopdtica. Buenos Aires: Editorial Albatros. Benson, T. (1986). Developing Tools for On Line Medical Reference Works. In R. Solomon, B. Blum and M. Jorgensen (Eds.), "MED1NFO 86." Amsterdam: Elsevier North Holland. Bratko, T. (1989). KARDIO: A study in deep and qualitative knowl- edge for expert systems. Cambridge, MA: The MIT Press. Clancey, W.J. (1981). NEOMYCIN: Reconfiguring a Rule-Based Expert System for Application to Teaching. In Proceedings of the Seventh IJCAI (pp. 829-836). Clancey, W.J., & Shortliffe, E. (Eds.). (1984). Readings in medical artificial intelligence. The first decade. Reading, MA: Addison- Wesley. Cooper, G. (1984). NESTOR, A Medical Decision Support System that Integrates Causal Temporal and Probabilistic Knowledge. PhD thesis, Medical Information Sciences, Stanford University, La Jolla, CA. Cristrbal, M,R., & Ortiz Latierro, N. (1991). Sistema Experto para el Tratamiento Homeopdtico del Glaucoma. MSc thesis, Facultad de lnformfitica--Universidad Politrcnica de Madrid, Madrid. Davis, R. (1976). Applications of metalevel knowledge to the con- struction, maintenance and use of large knowledge bases (AIM- 283). Computer Science Department, Stanford University. Fieschi, M. (1987). Inteligencia artificial en medicina. Barcelona: Masson S.A. Garcia (1993). C L A S E R Conjunto de Librer{as Abiertas para el De- sarrollo y Consulta de SS.EE. reutilizables. Final-year project. Facultad de lnformfitica, Madrid, Universidad Politrcnica de Madrid, Spain. Gascuel, O. (1981). SAM: Un systrme expert dans le domaine med- ical. In Proceedings of Congr~s AFCET RF-IA, Nancy, France. Amsterdam: North Holland. Gorry, G.A., Silverman, H., & Pauker, S.G. (1978). Capturing clinical expertise: A computer program that considers clinical response to digitalis. American Journal of Medicine, 64, 452-460. Hayes-Roth, F., Waterman, D.A., & Lenat, D.B. (1983). Building expert systems. Reading, MA: Addison-Wesley. Kulikowski, C.A., & Weiss, S.M. (1982). Representation of expert knowledge for consultation: The CASNET and expert projects. In P. Szolovits (Ed.), Artificial Intelligence in Medicine. Boulder, CO: Westview Press. Lathoud, J. (1988). Materia M~dica Homeopdtica. Buenos Aires: Editorial Albatros. Ledley, R., & Lusted, L. (1959). Reasoning formulations of medical diagnosis. Science 130, 9-24. Liebowitz, J., & De Salvo, D.A. (1989). Structuring expert systems: Domain, design and development. Englewood Cliffs, N J: Yourdon. Matr, J.L., & Pazos, J. (1988). Ingenier{a del conocimiento. Disoqo y Construccirn de sistemas expertos. Crrdoba, Argentina: SEPA. Miller, P. (1988). Selected topics in medical artfwial intelligence. New York: Springer Verlag. Pauker, S.G., & Szolovits, P. (1977). Analyzing and simulating taking the history of the present illness context formation. In Schneider and Sagwall Hein (Eds.), Computational linguistics in medicine. Amsterdam: North Holland. Pazos, J. (1989). Metodologfa construcci6n sistemas expertos. Notes MSc in Knowledge Engineering, Facultad de Informfitica-Univ- ersidad Potitrcnica de Madrid, Madrid. Pople, H. (1977). The formation of composite hypotheses in diagnostic problem solving. An exercise in synthetic reasoning. In Proceed- ings of the Fifth 1JCAL MIT, Cambridge, MA. Read, J., & Benson, T. (1986). Computer coding. British Journal of Healthcare Computing. Rubio, C. (1988). Indicaciones y contraindicaciones de la homeopat[a en oftalmolog{a. Notes C. Rubio, Asociaci6n de Home6patas de Madrid. Salatin, P., & Simonet, M. (1989). La medicina homeopathe. Gazette of the National Union of French Homeopathic Doctors. Shortliffe, E.M. (1976). Computer-based medical consultations: MY- CIN. New York: Elsevier. Shrogens, D. (1982). Specific and technical details about RADAR. Ghent, Belgium: ARCHIMEDE, Association pour la recherche en informatique medicale. Solomon, C., & Papert, S. (1976). A case study of a young child doing turtle graphics in LOGO. In Proceedings of the American Fed- eration of lnformation Processing Societies National Computer Conference. Vithoulkas, G. (1988). Vithoulkas expert system Universit6 de Na- mun, Institut d'lnformatique, Belguim. Waterman, B. (1986). A guide to expert systems. New York: Addison- Wesley.