Sistema Especialista em Medicina
/ Web-based diagnostic expert system
| developed by |
clinical domains |
keywords |
| Roberto Silva, Adventist University of São Paulo |
Multiple |
Expert systems, rule-based,
knowledge representation, diagnosis, advisor |
| status |
access demonstrator |
| Demonstrator developed
2005 |
|
| description |
Niaclin is a rule-based medical expert system designed to provide advice for medical symptoms.
The Niaclin knowledge base currently contains 116 diseases and
128 symptoms. The top ten
hyphoteses are listed for each query (allowing for differential diagnosis), and
links to these
principal hypotheses are provided.
Niaclin is a web-based version of Niacin,
a medical expert
system and expert
system shell
developed in 1995. The Niacin shell worked with several knowledge bases
including exanthematic diseases, causes of purpura, abdominal
pain, vaginitis.
Niaclin is based in
part on techniques used in Mycin (belief and disbelief measures),
Internist (descriptors associated with diseases
according to the frequency of the symptoms in diseases;
intrinsic value of symptoms) and in Addlasnig's CADIAG
(numeric values related to semantic variables).
Niaclin uses
six main rules to make a diagnosis and provide advice. The system takes account of:
- Symptoms present in a patient and described as related to a disease stored in the knowledge base
- Symptoms present in a patient but not specified as related to
disease specified in the knowledge base
- Symptoms not present in a patient but described as to a disease
specified in the knowledge base
- Symptoms not present in a patient and not linked to a
disease stored in the knowledge base
- Symptoms present in a patient but not included in a disease description in the knowledge base
- Symptoms not present in patient and not included in a disease description in the knowledge base.
Rules 1, 4 and 6 refer to positive findings and are used to increase the belief measure (MB).
Rules 2, 3 and 5 refer to negative findings and are used to decrease the belief measure (MD).
Confidence values are obtained by the difference between MB
and MD. The basic formula used for calculating the final confidence value in the system is:
MB = MB + [ ( 1- MB) * confirmability * intrinsic value of symptom)]
Confidence in belief measures increases
according to the frequency of symptoms linked to diseases,
the relative importance of symptoms,
the frequency of symptoms in the NIACLIN
knowledge base and may vary according
to the knowledge base rules. These and similar techniques are
used by several medical expert systems. With Niaclin, better
results are obtained when both positive and negative
findings observed in a patient are input in the
system.
|
| references |
Silva R, Parize MMG. NIACIN: Um Programa
para o Desenvolvimento de Sistemas Especialistas
em Medicina. Revista Informédica, 2 (11):
3-16, 1995. (Brasil)
[informaticamedica.org.br]
[]
|
" Desde o desenvolvimento
da Inteligência Artificial, têm surgido diversos
sistemas de apoio ao diagnóstico médico (SADMs),
voltados a várias especialidades, tais como
medicina interna, doenças bacterianas,
doenças renais, etc. Estes programas são
denominados de sistemas especialistas, ou
sistemas expert, e podem funcionar de diversas
maneiras, sendo a mais comum uma consulta
com base em questionamento do usuário. O mesmo
fornece sintomas e sinais do paciente, e o
programa elabora raciocínios lógicos e plausíveis
a partir dessa informação, fornecendo, então,
os prováveis diagnósticos. Esses raciocínios
automatizados utilizam uma base de conhecimentos
sobre a especialidade médica em questão, armazenada
previamente no computador. Na evolução da
Inteligência Artificial em Medicina, inicialmente
cada SADM correspondia a um programa especialmente
desenvolvido para este fim. Posteriormente,
entretanto, surgiram programas gerais, denominados
"shell", ou "vazios", que permitem se possa
trabalhar com diferentes bases de conhecimento,
usando um mesmo programa. O surgimento
de versões comercialmente disponíveis de programas
`shell" em Inteligência Artificial aumentou
significativamente o número de sistemas especialistas
desenvolvidos em Medicina, devido à facilidade
que os mesmos oferecem. "
|
|
| contact |
links |
Dr. Roberto Silva
Health Family Program
UNASP - Adventist University of São Paulo
Brazil
E: niaclin niaclin.com |
|
| acknowledgements |
| Dr. Roberto Silva,
Adventist University of São Paulo. |
Entry on OpenClinical: 06 May 2005
Last main update: 11 May 2005
|
|
Niaclin