Automatic Diagnosis of Mild Cognitive Impairment from Electroencephalogram Using Joint Wavelet Packet Decomposition and Common Spatial Pattern

Mild cognitive impairment (MCI) is identified as the initial stage of Alzheimer's disease. This condition presents less severe symptoms compared to Alzheimer's Disease (AD) to the extent that it does not significantly impact daily activities. Due to its subtle symptoms, diagnosing MCI is considerably more challenging than diagnosing Alzheimer's. However, early detection of MCI enhances the chances of treatment and prevention of its progression to Alzheimer's and dementia.

This study introduced a novel method for diagnosing MCI using an automated signal processing approach for electroencephalogram (EEG) signals. The method employs advanced signal processing techniques, including discrete wavelet transform in preprocessing and wavelet packet decomposition alongside spatial-spectral filters for feature extraction from EEG signals. EEG signals from 29 patients and 32 healthy individuals were utilized in this study.

The proposed method achieved a classification accuracy of 100% using a random subsampling validation approach. Wavelet packet decomposition effectively isolated frequency sub-bands within the EEG signals, enabling precise extraction. Furthermore, feature extraction using features extracted by the filter bank common spatial pattern (FBCSP) contributed to the increased classification accuracy of the two groups.

This study introduces a novel approach for MCI diagnosis by extracting spatial-spectral features from frequency sub-bands of EEG signals obtained through wavelet packet decomposition. The findings underscore the significance of wavelet packet decomposition in separating frequency sub-bands and applying a common spatial pattern filter on these sub-bands for effective feature extraction in distinguishing healthy individuals from those with MCI.