Improving Alzheimer's Disease Detection Efficiency with Scattering Wavelet Transform

Alzheimer Disease (AD) is a rapidly growing neurological disorder which have no cure. However early and accurate detection can help in better treatment planning to delay this process. EEG is a widely used modality which is completely non-invasive, portable and cost effective. In this study we used scattering wavelet transform (SWT) for decomposing the signal, obtained scattering coefficients as feature to train the machine learning model. Our sample size includes 80 AD subjects and 40 healthy subjects. We used 24 different classifiers to obtain the best classifier with highest accuracy in classification. Overall, we achieved 100% accuracy with 13 out of 19 channels and also with all 19 channels taking together. This shows the efficiency of SWT for feature extraction and obtaining drastic classification between AD and healthy subjects.