Estudo comparativo avaliando três modalidades de diagnóstico médico: parecer médico, buscas no Google e sistema especialista de apoio à decisão médica
Carregando...
Arquivos
Data
2020-02-21
Autores
Souza, Ademar Rosa de
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Estadual Paulista (Unesp)
Resumo
O conhecimento sobre qualquer patologia pode ser facilmente encontrado na internet, mas dificilmente encontra-se alguma ferramenta que faça a análise e o raciocínio entre os dados de um paciente e se obtenha o diagnóstico mais provável. Em nosso cotidiano, em virtude de uma maior demanda na área da saúde, existe uma necessidade crescente de diagnósticos médicos rápidos e precisos. Em virtude disso, foi elaborado um Sistema de Apoio à Decisão Médica com o intuito de otimizar e agilizar de forma confiável os diagnósticos médicos. A ideia é dar qualidade e agilidade à prática médica, adotando a tecnologia como ferramenta básica: “Quem tem mais informação, tem melhores condições para escolher e tomar decisões”. Na construção deste sistema, foram utilizados um banco de dados relacional (MySQL) e aplicadas técnicas de inteligência artificial, tais como: a construção de Árvores de Decisão, Aprendizado não supervisionado e a utilização das Redes de Bayes (onde estão envolvidos domínios de conhecimento com significativo grau de incerteza, como é o caso da área médica). Através da união destas técnicas, são feitas a seleção e classificação das doenças mais prováveis, onde as mesmas podem ser examinadas com mais detalhes pelo médico, garantindo assim uma maior segurança na escolha dos possíveis diagnósticos. Visando uma maior abrangência e rapidez na disseminação do conhecimento humano, o sistema foi disponibilizado via internet (www.danton.med.br). Para a concepção do projeto foi realizado um estudo prospectivo, randomizado, cruzado e aberto; no qual 3 grupos de médicos (denominados grupos: A, B e C – compostos por 4 médicos cada), realizaram diagnósticos de casos clínicos provenientes de um periódico médico, utilizando-se de 3 métodos distintos: Método 1 (Método tradicional; com parecer individual de cada médico, Método 2 (Utilização de ferramentas de buscas na internet - Google) e Método 3 (Utilização de um Sistema de Apoio à Decisão Médica). Para isso, foram selecionados 18 casos clínicos publicados no periódico médico “The New England Journal of Medicine”. Para análise do projeto foi utilizado o teste estatístico não paramétrico de Kruskal-Wallis, o qual permite trabalhar com amostras pequenas. Porém, contrariando as expectativas, no grupo avaliado (que incluía apenas os alunos do 5.º e 6.º ano da faculdade de medicina) não foi observada diferença estatística entre os três métodos utilizados (p=0,846). Concluiu-se, portanto, que o valor calculado se encontra na área de aceitação da hipótese nula, ou seja, aceita-se a hipótese de que as médias de todos os métodos (Parecer individual, Google e Sistema Especialista) não foram significativamente diferentes.
The knowledge about any pathology can be easily found on the internet, but it is difficult to find any tool that makes the analysis and reasoning between the data of a patient and obtain the most probable diagnosis. In our daily lives, due to a greater demand in the health area, there is a growing need for fast and accurate medical diagnoses. As a result, a Medical Decision Support System was developed in order to reliably optimize and streamline medical diagnostics. The idea is to give quality and agility to medical practice, adopting technology as a basic tool: “Who has more information, has better conditions to choose and make decisions”. In the construction of this system, a relational database (MySQL) was used and artificial intelligence techniques were applied, such as: the construction of Decision Trees, Unsupervised Learning and the use of Bayes Networks (where knowledge domains are involved with significant degree of uncertainty, as is the case in the medical field). Through the union of these techniques, the selection and classification of the most probable diseases are made, where they can be examined in more detail by the doctor, thus ensuring greater security in the choice of possible diagnoses. Aiming at a greater scope and speed in the dissemination of human knowledge, the system was made available via internet (www.danton.med.br). To design the project, a prospective, randomized, crossover and open study was carried out; in which 3 groups of doctors (called groups: A, B and C - composed of 4 doctors each), made diagnoses of clinical cases from a medical journal, using 3 different methods: Method 1 (Traditional method; with individual opinion from each doctor, Method 2 (Use of internet search tools - Google) and Method 3 (Use of a Medical Decision Support System). For this, 18 published clinical cases were selected in the medical journal “The New England Journal of Medicine”. The Kruskal-Wallis non-parametric statistical test was used to analyze the project, which allows working with small samples. However, contrary to expectations, in the evaluated group (which included only 5th and 6th year medical school students), no statistical difference was observed between the three methods used (p=0.846). It was concluded, therefore, that the calculated value is in the area of acceptance of the null hypothesis, that is, it is accepted the hypothesis that the averages of all methods (Individual opinion, Google and Expert System) were not significantly different.
The knowledge about any pathology can be easily found on the internet, but it is difficult to find any tool that makes the analysis and reasoning between the data of a patient and obtain the most probable diagnosis. In our daily lives, due to a greater demand in the health area, there is a growing need for fast and accurate medical diagnoses. As a result, a Medical Decision Support System was developed in order to reliably optimize and streamline medical diagnostics. The idea is to give quality and agility to medical practice, adopting technology as a basic tool: “Who has more information, has better conditions to choose and make decisions”. In the construction of this system, a relational database (MySQL) was used and artificial intelligence techniques were applied, such as: the construction of Decision Trees, Unsupervised Learning and the use of Bayes Networks (where knowledge domains are involved with significant degree of uncertainty, as is the case in the medical field). Through the union of these techniques, the selection and classification of the most probable diseases are made, where they can be examined in more detail by the doctor, thus ensuring greater security in the choice of possible diagnoses. Aiming at a greater scope and speed in the dissemination of human knowledge, the system was made available via internet (www.danton.med.br). To design the project, a prospective, randomized, crossover and open study was carried out; in which 3 groups of doctors (called groups: A, B and C - composed of 4 doctors each), made diagnoses of clinical cases from a medical journal, using 3 different methods: Method 1 (Traditional method; with individual opinion from each doctor, Method 2 (Use of internet search tools - Google) and Method 3 (Use of a Medical Decision Support System). For this, 18 published clinical cases were selected in the medical journal “The New England Journal of Medicine”. The Kruskal-Wallis non-parametric statistical test was used to analyze the project, which allows working with small samples. However, contrary to expectations, in the evaluated group (which included only 5th and 6th year medical school students), no statistical difference was observed between the three methods used (p=0.846). It was concluded, therefore, that the calculated value is in the area of acceptance of the null hypothesis, that is, it is accepted the hypothesis that the averages of all methods (Individual opinion, Google and Expert System) were not significantly different.
Descrição
Palavras-chave
Medicina, Inteligência artificial, Sistema especialista de apoio à decisão médica, Redes de Bayes, Árvores de decisão, Medicine, Artificial intelligence, Specialist medical decision support system, Bayes networks, Decision trees