Machine Learning-Based Detection of Depression from EEG Signals Using Linear and Non-Linear Features

Non-linear, complex, and non-stationary signals are EEGs. Deterioration in brain function is shown in electroencephalogram (EEG) data throughout MDD or despair. Dementia patients and healthy people were classified using linear and non-linear variables, as well as a combination of the two, in this research. All of the data utilized in this research came from freely accessible public sources. There were 30 healthy persons and 34 patients with MDD in the sample. These classifiers were utilized: quadratic discriminant analysis, linear discriminant analysis, radial basis function network, and multi layered perceptron neural network (MLPNN). With the usage of linear features, the alpha power with MLPNN classifier achieved the greatest classification accuracy of 92.58 percent. Using a non-linear feature, RWE and WE both accomplished 90.67% classification accuracy with the RBFN classifier and the LDA classifier, respectively. When using a mix of linear and non-linear features, such as alpha power and RWE with MLPNN and RBFN classifier, the best classification accuracy was 95.3%. The maximum classification accuracy was 95.2%, which was accomplished by combining non-linear characteristics, namely RWE and WE. This was shown in the study as well. Researchers looked examined how various classifiers performed when given either linear or non-linear characteristics, or a mix of the two, as well as their sensitivity and specificity. Alpha power and RWE together had the best classification accuracy (93.33%) among all classifiers tested.