A Study of EEG Signal Classification in Alzheimer's Disease Using the IMPSO for Channel Selection

Alzheimer's disease (AD) is a neurodegenerative disease that is primarily characterised by insidious onset and subtle clinical symptoms, which can make it challenging to diagnose through unconventional means. EEG has been extensively employed as a primary diagnostic tool for Alzheimer's disease (AD), facilitating the assessment and diagnosis of various neurological disorders. An accurate diagnosis is crucial for the prevention of the transition from early cognitive impairment to Alzheimer's disease (AD) and the provision of early treatment for AD patients. The maj ority of classification methods concentrate solely on the time or frequency domain, thereby neglecting the interconnectivity between the two. Accordingly, this paper generates time- frequency spectrograms based on a wavelet transform for analysis. However, the time-frequency spectrograms contain a considerable amount of redundant information, which results in a sudden increase in computation. To address this issue, this paper proposes a bidirectional principal component analysis (BDPCA) method to reduce the dimensionality of the time- frequency spectrogram, thereby reducing the computational burden while effectively restoring the features of the image. Subsequently, the improved PSO algorithm was employed to select the 16 channels, and seven key channels (Fp1, F7, F8, F3, T3, T4 and 01) were ultimately identified. The features of these selected channels were then inputted into the XGBoost classifier to achieve the recognition of patients and normal controls, with an accuracy rate of 93.56%. This study is the inaugural application of the IMPSO optimisation algorithm and time- frequency spectrogram to EEG channel selection, thereby establishing an efficient framework for AD detection. The method reduces the computational complexity and improves the generalisation ability, which facilitates the rapid identification and reduction of misdiagnosis in a clinical setting while also advancing the understanding of the brain function mechanisms underlying neurological diseases.