Loading...
Publication
Índice de Severidade para a Doença de Parkinson baseado em Redes Neuronais Artificiais
| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 2.42 MB | Adobe PDF |
Authors
Abstract(s)
A Doença de Parkinson (DP) é uma condição neurodegenerativa progressiva que afeta,
principalmente, o sistema motor das pessoas. A prevalência da DP tem aumentado
significativamente, afetando 1% da população acima de 60 anos e até 4% da população
acima de 80 anos. Um índice com intuito de verificar a severidade dessa doença para
acompanhar ou até mesmo diagnosticar precocemente os pacientes, é fundamental. Este
trabalho propõe o desenvolvimento de um índice de severidade para classificar os
pacientes, a partir de sintomas, principalmente motores, utilizando uma Rede Neuronal
Artificial (RNA), treinada pelo algoritmo Self-Organizing Maps (SOM). O SOM, com
base na similaridade das informações, é capaz de agrupar os dados num mapa
bidimensional. Além disso, utilizou-se o algoritmo K-means para associar esses dados
em Classes. Foi utilizado o banco de dados FOX Insight, que oferece dados em forma de
questionários respondidos por pacientes ou cuidadores de todo o mundo, com uma
variedade de informações referente a esta patologia. Após o pré-processamento dos
dados, selecionou-se um conjunto de 597 questionários (pacientes), contendo 28
perguntas definidas. Os dados foram codificados de acordo com cada grupo de
perguntas. Os sintomas foram individualmente analisados após o mapeamento e
divididos em quatro Classes. Na Classe 1, a maioria dos sintomas não estavam presentes.
Na Classe 2, a presença de determinados sintomas demonstraram marcos iniciais da
doença, como movimentos mais lentos, voz mais suave e tremores. Na Classe 3, os
sintomas relacionados as mobilidades do paciente, em particular, a dor, destacam-se
entre os mais relatados. Na Classe 4, observa-se a presença intensa de todos os sintomas,
sendo os mais frequentes, considerados sintomas em uma fase mais avançada da doença.
Dessa forma, a Classe 1 representa pacientes assintomáticos ou em uma fase muito leve,
a Classe 2, refere-se a uma fase leve, a Classe 3, uma fase moderada da doença e a Classe
4 uma fase grave. Para testar a ferramenta, foram utilizados dados de alguns desses
pacientes, que responderam ao mesmo questionário outras vezes em momentos
espaçados (simulando as consultas médicas). Por fim, foi possível identificar o estágio
atual da doença possibilitando acompanhá-los, sendo essa uma ótima ferramenta de
suporte à decisão.
Parkinson's disease (PD) is a progressive neurodegenerative condition affecting the motor system. With advancing age, the prevalence of PD has increased significantly affecting 1% of the population over 60 years and up to 4% of the population over 80 years. This paper proposes the development of a severity index to classify patients possibly suffering from PD. To achieve this goal, questions were selected from a platform that offers approximately 53,000 data in questionnaires provided by patients or caregivers around the world with a wide range of information available, referring to this pathology. In which, after all the cleaning of the data obtained, a set of 597 individual data remained. These data were numerically encoded for the purpose of building a self-organizing map (SOM), which maps data onto a grid of artificial neurons. The methodology used generated four Classes of symptom severity, these being subdivided into: mild (Class 1), mild (Class 2), moderate (Class 3), and severe (Class 4), which are composed of 27 symptoms defined at the beginning of this paper. In Class 1 less than 1% of the patients have existing symptoms. In Class 2, the symptoms are more recurrent, signaling the onset of symptoms, followed by mild signs of difficulty in moving, swallowing, as well as a slight alteration in speech. However, in Class 3, the symptoms related to the patient's mobility stand out among the most reported. Class 4 shows the presence of all the symptoms selected in patients in general, being considered more severe and related to independence. For the validation we used the complete data offered by the platform, in which the proposed severity scores not only identify the current stage of a patient's disease, but also offer the physician an easy-to-read 2D map that makes it possible to follow the progression of the disease.
Parkinson's disease (PD) is a progressive neurodegenerative condition affecting the motor system. With advancing age, the prevalence of PD has increased significantly affecting 1% of the population over 60 years and up to 4% of the population over 80 years. This paper proposes the development of a severity index to classify patients possibly suffering from PD. To achieve this goal, questions were selected from a platform that offers approximately 53,000 data in questionnaires provided by patients or caregivers around the world with a wide range of information available, referring to this pathology. In which, after all the cleaning of the data obtained, a set of 597 individual data remained. These data were numerically encoded for the purpose of building a self-organizing map (SOM), which maps data onto a grid of artificial neurons. The methodology used generated four Classes of symptom severity, these being subdivided into: mild (Class 1), mild (Class 2), moderate (Class 3), and severe (Class 4), which are composed of 27 symptoms defined at the beginning of this paper. In Class 1 less than 1% of the patients have existing symptoms. In Class 2, the symptoms are more recurrent, signaling the onset of symptoms, followed by mild signs of difficulty in moving, swallowing, as well as a slight alteration in speech. However, in Class 3, the symptoms related to the patient's mobility stand out among the most reported. Class 4 shows the presence of all the symptoms selected in patients in general, being considered more severe and related to independence. For the validation we used the complete data offered by the platform, in which the proposed severity scores not only identify the current stage of a patient's disease, but also offer the physician an easy-to-read 2D map that makes it possible to follow the progression of the disease.
Description
Keywords
Doença de Parkinson Índice de Severidade Redes Neuronais Artificiais Self-Organizing Maps