Name: | Description: | Size: | Format: | |
---|---|---|---|---|
742.17 KB | Adobe PDF |
Advisor(s)
Abstract(s)
Este trabalho tem como objetivo a detecção precoce das doenças de Alzheimer e Parkinson através de parâmetros não-lineares multibanda de sinais EEG. Para cada par de grupos de estudo, uma seleção dos parâmetros é realizada através de algoritmo genético. Os parâmetros selecionados são utilizados como entrada para classificadores com validação cruza da leave-one-out. Acurácias de classificação de 100% são obtidas, empelo menos uma sub-banda, para 3 pares de grupos de estudo enquanto 90,60% é alcançado para o par Controle vs Alzhei-mer/Parkinson. A sub-banda delta foi a que, em geral, apresentou maiores diferenças significativas entre os grupos.
This work aims to detect Alzheimer’s and Parkinson’s diseases at early stage through non-linear multiband para-meters of EEG signals. For each pair of study groups, parameters selection was performed through genetic algorithm. The selected parameters are used as input for classifiers with leave-one-outcross-validation. Classification accuracies of 100% are achieved, in at least one sub band, for 3 pairs of study groups while90.60% is achieved for the Control vs Alzheimer/Parkinson pair. The delta sub band showed, in general, the greatest significant differences between the groups.
This work aims to detect Alzheimer’s and Parkinson’s diseases at early stage through non-linear multiband para-meters of EEG signals. For each pair of study groups, parameters selection was performed through genetic algorithm. The selected parameters are used as input for classifiers with leave-one-outcross-validation. Classification accuracies of 100% are achieved, in at least one sub band, for 3 pairs of study groups while90.60% is achieved for the Control vs Alzheimer/Parkinson pair. The delta sub band showed, in general, the greatest significant differences between the groups.
Description
Keywords
Detecção precoce Alzheimer Parkinson EEG Análise não-linear Wavelet Early detection Nonlinear analysis